Time Series Resources for teachers
advertisement
Time Series Resource for Secondary School Statistics 1 Crown copyright © This work is licensed under the Creative Commons Attribution 3.0 New Zealand licence. You are free to copy, distribute, and adapt the work, as long as you attribute the work to Statistics NZ and abide by the other licence terms. Please note you may not use any departmental or governmental emblem, logo, or coat of arms in any way that infringes any provision of the Flags, Emblems, and Names Protection Act 1981. Use the wording ‘Statistics New Zealand’ in your attribution, not the Statistics NZ logo. Liability While all care and diligence has been used in processing, analysing, and extracting data and information in this publication, Statistics New Zealand gives no warranty it is error free and will not be liable for any loss or damage suffered by the use directly, or indirectly, of the information in this publication. Citation Statistics New Zealand (2015).Time Series: Resource for NZQA. Wellington: Statistics New Zealand Published in July 2015 Statistics New Zealand Tatauranga Aotearoa Wellington, New Zealand Contact Statistics New Zealand Information Centre: info@stats.govt.nz Phone toll-free 0508 525 525 Phone international +64 4 931 4600 www.stats.govt.nz 2 INDEX Introduction ........................................................................ 4 Purpose ............................................................................ 4 Summary ......................................................................... 4 Graphs ............................................................................. 4 Limitations of this resource .............................................. 4 Time Series Datasets ........................................................... 5 1. Average number of visitors arrivals by purpose- Multiplicative model 5 2. Number of guest nights by type of accommodation-Multiplicative model 17 3. Litres (in millions) of alcohol available for consumption-Multiplicative model 29 4. Deaths by regional councils of the total population- Multiplicative model 42 5. Number of pigs killed in New Zealand- Multiplicative model 54 6. Gross weight (tonnes) of imports by seaports in New Zealand- Multiplicative model 66 7. Persons employed in Labour Force Aged 15-19 and 20-24- Multiplicative model 8. Number of flights by direction for Buenos Aires – Additive model Acknowledgements.......................................................... 103 3 90 79 Introduction Purpose The purpose of this document is to provide teachers with a new resource in order to describe trends in real-context time series data. It has been created in order to cover some of the material in Level 3 NCEA Achievement Standards such as: Investigate time series data- 91580 Use statistical methods to make a formal inference- 91582 Summary Eight monthly and quarterly time series datasets have been collected from Statistics New Zealand and have been customised in “Infoshare” to CSV files appropriate/adequate for iNZight. Seven are multiplicative and one is additive. For each dataset the total data for the variables being analysed have been added. Model paragraphs as to how a professional statistician would describe the trend, seasonal, and remainder of the variables being compared have been added. There is also a paragraph that highlights comparisons and differences between the variables being analysed and the total data. At the beginning of each time series resource the “Total” data for the variables that are being analysed have been included with no commentary. It has been used to see whether the variables that have been analysed have had an effect on the total. In addition, this enables users to provide their own commentaries about the “Total” data to test their knowledge. At the end of each time series resource a set of notes on how to obtain the particular dataset from the Statistics New Zealand website as well as how to import and analyse the data in iNZight has been added. Also, for some of the time series datasets, a “useful links” section in the metadata is included whereby references have been used that have been used in model answers. Graphs All components in the decomposition graph are plotted on the same scales. This is helpful when looking at the magnitude differences between components, especially when deciding which component (seasonal or irregular) contributes more of the variability in the series. Limitations of this resource This resource does not include forecasting. Most of Statistics New Zealand's economic series are multiplicatively adjusted, in line with standard practice among producers of seasonally adjusted data, which is to assume that the components of an economic time series are multiplicatively adjusted. All these examples are made using Official Statistics. Therefore, spikes in the irregular component are likely to be due to real world events rather than untreated errors or extreme observations. Details of how errors and extreme observations are treated can be found in the data quality section of the relevant media release, which can be found by using the link provided in the “How to access dataset” section under each example provided in this resource. All datasets have been created using version 2.0.4 of iNZight. 4 Time Series Datasets 1. Average number of visitors arrivals by purpose- Multiplicative model Total data for visitor arrivals by all purposes: Time series plot for Total.for.all.travel.purposes 2e+05 150000 1e+05 2000 2005 2010 Time 5 2015 Decomposition of data:Total.for.all.travel.purposes 250000 Trend Raw data 2e+05 150000 1e+05 50000 Seasonal Swing 60000 40000 20000 0 -20000 4000 2000 0 -2000 -4000 -6000 Residuals 2000 2005 Time 6 2010 2015 Seasonal plot for Total.for.all.travel.purposes Multiplicative seasonal effects 1.4 2015 2014 2002 2013 2011 2001 2009 2010 2006 2008 2012 2007 2003 2002 2005 2004 2000 1.1 150000 1.2 1.3 2008 2010 2011 2007 2013 2012 2009 2006 2005 2004 2003 1.0 200000 2014 2001 1999 100000 2000 0.9 1999 0.8 1998 Jan - Mar Apr - Jun Jul - Sep Oct - Dec Jan - Mar Quarter Apr - Jun Jul - Sep Quarter 7 Oct - Dec Visitor arrivals – Holiday/Vacation purpose The trend for this series, illurstrated in the time series plot and the Decomposition graph, has been gradually increasing since 1998. The number of visitor arrivals for holiday purposes has increased from around 35,000 in 1998 to 60,000 in 2015. However, it should be noted that a slight dip in 2012 is evident. This could possibly indicate an event or a problem in the data set, therefore, further investigation is needed. The detail of the seasonal pattern is easily read in the “Multiplicative Seasonal Effects” graph which shows a very regular seasonal pattern with highs and lows in the same quarters each year with visitor numbers high in the first quarter of each year (January until March) and low in the third quarter (July- September). In addition, the graph also illurstrates a significant increase in travel due to holiday purposes in the fourth quarter, increasing further in the first quarter of the following year.This is most likely due to the summer season. Looking at the seasonal pattern in the Decomposition graph, the amplitude seems to be increasing, suggesting that the increase of visitors coming to New Zealand for holiday purposes is becoming more seasonal. In addition, the residuals in the Decomposition shows that the irregular has no particular pattern and varies between 1000 and -3000. Looking at the minimum and maximum, the minimum residual in 2008 could possibly suggest that visitor arrivals for holiday purposes decreased due to the Global Financial Crisis. In addition, the maximum residual in 2011 could possibly be due to the 2011 Rugby World Cup. The variation of the irregular is small compared to the trend and seasonal component. Time series plot for Holiday.Vacation 1e+05 80000 60000 40000 2000 2005 2010 Time 8 2015 Decomposition of data:Holiday.Vacation Trend Raw data 1e+05 80000 60000 40000 20000 40000 Seasonal Swing 30000 20000 10000 0 -10000 2000 1000 0 -1000 -2000 -3000 Residuals 2000 2005 Time 9 2010 2015 Multiplicative seasonal effects 1.6 2015 2014 2007 2008 2010 2011 2004 2005 2006 2003 2009 2013 2012 1.4 8e+04 1e+05 Seasonal plot for Holiday.Vacation 2002 2001 6e+04 1.2 2014 2003 2006 2013 2011 2009 2004 2010 2002 2007 2005 2008 2000 1999 2012 1.0 2001 2000 4e+04 1999 0.8 1998 Jan - Mar Apr - Jun Jul - Sep Quarter Oct - Dec Jan - Mar 10 Apr - Jun Jul - Sep Quarter Oct - Dec Visitor arrivals – Conference/Convention purpose Looking at the trend in the time series plot and the Decomposition graph, it shows a relatively steep increase from 800 to 1250 in visitor arrivals for conference/convention purposes between 2004 and 2006 before leveling off and decreasing to 1000. This may be due to international conferences possibly being held in New Zealand during 20042006 and then being moved elsewhere, for example the 7th World Plumbing Conference in Auckland 2005 and the 9th Annual AVAR International Conference in Auckland 2006. Referring to the Multiplicative Seasonal Effects graph, the seasonality shows that the number of vistors arriving for conference purposes are high in quarters 1 and 4 (OctMarch) and are low in quarter 2 (April-June). The seasonal pattern in the Decomposition graph shows a relatively constant amplitude, varying from 100 to -200. It is small compared to the trend, however it is relatively similar to the seasonal component. Close examination of the irregular pattern in the Residual section of the Decomposition graph shows some highs and lows varying from 200 to -100, which could possibly be due to conferences/conventions being held every two-five years or large-one off conferences, however further investigation will be needed to validate this. Time series plot for Conventions.Conferences 1600 1400 1200 1000 800 2000 2005 2010 Time 11 2015 Decomposition of data:Conventions.Conferences Trend Raw data 1600 1400 1200 1000 800 600 Seasonal Swing 100 50 0 -50 -100 -150 -200 Residuals 200 100 0 -100 2000 2005 Time 12 2010 2015 Multiplicative seasonal effects 1600 Seasonal plot for Conventions.Conferences 1.10 2005 2008 1.05 1400 2004 2013 2006 2004 2014 2011 2005 2007 2010 1.00 1200 2010 2013 2003 2009 2006 2011 2003 2012 2015 0.95 2014 1000 2001 2001 2009 2007 2008 2002 0.90 2000 1999 2012 2000 800 1998 2002 0.85 1999 Jan - Mar Apr - Jun Jul - Sep Oct - Dec Jan - Mar Quarter Apr - Jun Jul - Sep Quarter 13 Oct - Dec COMPARISON OF THE SERIES The actual counts of visitor arrivals for holiday purposes are greatly higher than those visiting for conference/convention purposes. For example the highest value for holiday purposes is around 100,000 whereas the highest value for conferences is around 1580. The two series display distinct patterns in the trend, with visitor arrivals for holiday purposes showing a gradual increase from 3000 to 5000, compared to visitors for conference purposes, showing a significant rise of 400 people between 2004-2006 before levelling off and gradually decreasing by 200. Both series show distinct but different seasonal patterns. The seasonal data for holiday/vacation purposes reflects the total data for all travel purposes exactly, indicating that the total data is largely influenced by the holiday/vacation counts. In comparison, the seasonal component for conference purposes shows different seasonality compared to the total data for all travel purposes. The seasonal component for conference/convention data highlights a significant of decrease from quarter 1 to 2, then a significant increase from quarter 2 to 3. It is likely that due to the small counts of visitor arrival for conference/convention purposes, this seasonality is masked by the other variables when compared to the total data for all travel purposes. The irregulars for visitor arrivals for holiday purposes shows no particular pattern. Conversely, the irregulars for visitor arrivals for conference purposes shows some highs and lows. The decomposition of both the holiday and conference purpose data when compared to the all purposes data, all indicate a minimum residual around 2008. This could possibly be due to the Global Financial Crisis, hence the reason for less international visitors for whatever purpose. ACCESSING THE DATA How do I access the International Travel and Migration Statistics? This series is found in the International Travel and Migration section of the Statistics New Zealand website: www.stats.govt.nz (Statistics NZ Home > Browse for statistics > Population > Migration > International Travel and Migration) How do I get this particular data set and upload it to iNZight? On the Statistics New Zealand website you will find Quick links and Infoshare (www.stats.govt.nz/infoshare). Use the Browse tab. Select Tourism > International Travel and Migration-ITM > Average number of visitors in New Zealand by purpose (Qrtly-Mar/Jun/Sep/Dec). Under Travel Purpose, hold Ctrl and select: Conventions/Conferences Holiday/ Vacation TOTAL ALL TRAVEL PURPOSES Select all the Time periods. At the bottom right of the screen, change Table on screen to csv file. 14 Read this csv file into iNZight. You will need to remove unnecessary information from the csv first and rename the variables so that only the first row of the csv contains variable names. Useful links: Global Financial Crisis- http://www.globalissues.org/article/768/global-financial-crisis Rugby World Cup 2011- http://www.rugbyworldcup.com/ Convention Management NZ- http://www.cmnzl.co.nz/default,events,,past.sm How to produce Time Series graph, Decomposition graph and Seasonal plot in iNZight? After you have imported your Time Series data click “Advanced” and then “Time Series…” The following command window will appear: Click if your dataset is Additive or Multiplicative. Then select the Time Series Plot option to produce the TOTAL graph, which includes the trend line. Then decompose the dataset using the Decompose option. This will produce the “Decomposition Total” graph. Then select the Seasonplot option to produce the Seasonal plot and Multiplicative/Additive Seasonal Effects. 15 How do you format your time variable so iNZight recognises it? INZight assumes that datasets are quarterly when producing a time series. If your time variable is not in the format the program recognises or if you have not included a quarter column, then you can supply the time structure-information through the module's command window. 1. Click Provide Time Information and then press Create. The following pop-up box will appear: 2. In Start Date enter the year in which the series started 3. In Frequency enter the number of seasons in a year (e.g. 52 for weekly data) 4. In Season Number enter the number of the season at which the series started (e.g. 3 for 3rd week of the year) Reinterpret the above if your data runs over weeks not years and the seasons are days of the week, or your data concerns days and hours in the day. 16 2. Number of guest nights by type of accommodation-Multiplicative model Total number of guest nights for all accommodation types: Time series plot for Total.guest.nights.by.all.Accomodation.Types 4500000 4e+06 3500000 3e+06 2500000 2e+06 1500000 1e+06 2000 2005 2010 Time 17 2015 Decomposition of data:Total.guest.nights.by.all.Accomodation.Types 5e+06 Trend Raw data 4e+06 3e+06 2e+06 1e+06 Seasonal Swing 2e+06 1500000 1e+06 5e+05 0 -5e+05 3e+05 2e+05 1e+05 0 -1e+05 -2e+05 -3e+05 -1e+06 Residuals 2000 2005 Time 18 2010 2015 Multiplicative seasonal effects 2015 1.6 2010 2008 2014 2011 2007 2005 2009 2006 2012 2013 2004 2003 2002 2001 2000 1999 1998 1.4 2014 2013 2012 2009 2011 2007 2010 2006 2004 2008 2003 2005 2002 2001 1997 1.2 3000000 3500000 4000000 4500000 Seasonal plot for Total.guest.nights.by.all.Acommodation.Types 2500000 2000 1000000 0.8 1500000 2000000 1.0 1999 1998 1997 1996 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Month Feb Mar Apr May Jun Jul Month 19 Aug Sep Oct Nov Dec Number of guest nights - Hotels: The trend for Hotels, illustrated on the time series plot and the Decomposition graph for Hotels, shows an increase from 550,000 in 1996 to 1,000,000 in 2015. However, it is evident that are two dips within the trend line around 1998 and 2012. In 2011 there was a local peak (possibly due to the Rugby World Cup), then a decrease, but the trend has been increasing steadily since then. The detail of the seasonal pattern can be read in the Multiplicative Seasonal Effects graph, where guest nights in hotels are highest between February and March and lowest in June each year. There is a steep increase in July, then the seasonal component flattens until September. From October it rises steeply to November. There is a steep decrease in December, then a sharp increase to January. This may be due to the summer holiday season, as many people travel across the country around this time of year. Looking at the seasonal pattern in the Decomposition graph, the amplitude seems to be increasing since 1996, suggesting an increase in the number of guest nights in hotels across New Zealand, varying from 200,000 to -200,000. In addition, the residuals in the Decomposition shows that the irregular has no particular pattern and has varied between 50000 and -100,000 since 1996. It small compared to the trend and seasonal component. Looking at the minimum residual in 2011 could possibly suggest that number of guest nights in hotels decreased mainly due to earthquakes in Canterbury. It is small compared to the seasonal component and the trend. Time series plot for Hotels 1200000 1e+06 8e+05 6e+05 4e+05 2000 2005 2010 Time 20 2015 Decomposition of data:Hotels Trend Raw data 1200000 1e+06 8e+05 6e+05 4e+05 Seasonal Swing 2e+05 1e+05 0 -1e+05 -2e+05 50000 Residuals 0 -50000 -1e+05 2000 2005 Time 21 2010 2015 Multiplicative seasonal effects 1.2 2015 1000000 2010 2013 2008 2012 2007 2009 2005 1.1 2014 2014 2011 2013 2012 2010 2011 2009 2006 2004 2003 2006 2007 2008 2004 2003 2005 2002 800000 2002 2001 1.0 1200000 Seasonal plot for Hotels 2001 2000 0.9 2000 600000 1999 1997 1999 1996 1998 1998 400000 0.8 1997 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Month Feb Mar Apr May Jun Jul Month 22 Aug Sep Oct Nov Dec Number of guest nights - Holiday Parks: The trend for this series, illustrated in the time series plot and the Decomposition graph, has remained relatively constant since 1997. There is a slight increase from about 400,000 to 500,000. The detail of the seasonal pattern can be read in the Multiplicative Seasonal Effects graph, which shows visitor numbers are highest in January. The size of the seasonal component decreases throughout the year (although it slows down around March-April which could be due to Easter holidays) with a trough in June-Aug, then a small rise in July possibly caused by school holidays. After August, the size of the seasonal component rises steadily, slowing a little in November then rising sharply in December. There is another sharp rise in January, probably due to domestic summer holidays and overseas tourists. Many families, schools and sports teams choose to stay in holiday parks because they are cheaper and accommodate for larger groups. Looking at the seasonal pattern in the Decomposition graph, the amplitude seem relatively constant, suggesting the number of guest nights in holiday parks do not vary significantly. The seasonal plot for holiday parks shows very little variation year-on-year. There is more variation in the summer months (Nov-April) than the winter months. The variation in the March and April values may be due to Easter being in either of these 2 months. In addition, the residuals in the Decomposition shows that the irregular has no particular pattern and varies between 200,000 and -200,000. It is small compared to the trend and seasonal component. Looking at the maximum residual in 1998, could possibly suggest that number of guest nights in holiday parks increased due to a particular event which needs further investigation. Time series plot for Holiday.Parks 1500000 1e+06 5e+05 2000 2005 2010 Time 23 2015 Decomposition of data:Holiday.Parks Trend Raw data 1500000 1e+06 5e+05 0 Seasonal Swing 1e+06 5e+05 0 2e+05 1e+05 0 -1e+05 -2e+05 Residuals 2000 2005 Time 24 2010 2015 Multiplicative seasonal effects 2010 2008 2005 2009 2004 2015 2011 2007 2006 1998 2002 2014 1999 2001 2012 2003 2000 2013 3.5 1500000 Seasonal plot for Holiday.parks 1000000 2.5 3.0 1997 0.5 1.0 500000 1.5 2.0 2014 2013 2009 2012 2011 2001 2007 2003 2004 2008 2002 2010 2006 2000 2005 1997 1999 1998 1996 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Month Feb Mar Apr May Jun Jul Month 25 Aug Sep Oct Nov Dec COMPARISON OF THE TWO SERIES The variation in the number of actual guest nights for holiday parks is comparatively larger than hotels. The number of guest nights for holiday parks varies between 100,000 and 1,500,000 whereas the number of guest nights for hotels varies between 300,000 and 1,200,000. The two series display distinct patterns in the trend, with number of guest nights in hotels showing an increase from 550,000 to 1,000,000, compared to the number of guest nights in holiday parks, which remains relatively constant around 500,000. These two trend lines indicate that the number of guest nights in hotels are relatively consistent throughout most months of the year, and that the number of guest nights in holiday parks undergoes a boom and bust, with a significant increase in the number of guest nights in the summer school holiday season (January) and then relatively low numbers throughout the rest of the year. Both series show distinct but different seasonal patterns. The seasonal component for number of guest nights in hotels shows different seasonality compared to the overall data for all accommodation types. The seasonal component for number of guest nights in hotels highlights a relative increase of 0.2 from June to July, then another increase of 0.38 from September through to November. It is likely that due to the lower counts of guest nights in hotels, the seasonality is masked by the other variables when compared to the overall accommodation type data. The irregulars for hotels and holiday parks show no particular pattern and have smaller ranges when compared to their corresponding trends and seasonal components. ACCESSING THE DATA How do I access the Accommodation Survey Statistics? This series is found in the Accommodation section of the Statistics New Zealand website: www.stats.govt.nz (Statistics NZ Home > Browse for statistics > Industry sectors > Accommodation > Accommodation Survey) How do I get this particular data set and upload it to iNZight? On the Statistics New Zealand website you will find Quick links and Infoshare (www.stats.govt.nz/infoshare). Use the Browse tab. Select Tourism > Accommodation Survey - ACS > Actual by Accommodation by Type by Variable (Monthly). Under Accommodation Survey Actual by Accommodation Type 2009 Required variable, hold Ctrl and select: Hotels Holiday Parks Total Select all the Time periods. 26 At the bottom right of the screen, change Table on screen to csv file. Read this csv file into iNZight. You will need to remove unnecessary information from the csv first and rename the variables so that only the first row of the csv contains variable names. Useful links: Story: Historic earthquakes Page 12 – The 2010 Canterbury (Darfield) earthquake- http://www.teara.govt.nz/en/historic-earthquakes/page-12 How to produce Time Series graph, Decomposition graph and Seasonal plot in iNZight? After you have imported your Time Series data click “Advanced” and then “Time Series…” The following command window will appear: Click if your dataset is Additive or Multiplicative. Then select the Time Series Plot option to produce the TOTAL graph, which includes the trend line. Then decompose the dataset using the Decompose option. This will produce the “Decomposition Total” graph. Then select the Seasonplot option to produce the Seasonal plot and Multiplicative/Additive Seasonal Effects. 27 How do you format your time variable so iNZight recognises it? INZight assumes that datasets are quarterly when producing a time series. If your time variable is not in the format the program recognises or if you have not included a quarter column, then you can supply the time structure-information through the module's command window. 1. Click Provide Time Information and then press Create. The following pop-up box will appear: 2. In Start Date enter the year in which the series started 3. In Frequency enter the number of seasons in a year (e.g. 52 for weekly data) 4. In Season Number enter the number of the season at which the series started (e.g. 3 for 3rd week of the year) Reinterpret the above if your data runs over weeks not years and the seasons are days of the week, or your data concerns days and hours in the day. 28 3. Litres (in millions) of alcohol available for consumption-Multiplicative model Total volume of All Beverages available for consumption: Time series plot for All.Beverages 11 10 9 8 7 6 5 1985 1990 1995 2000 Time 29 2005 2010 2015 Decomposition of data:All.Beverages Trend Raw data 11 10 9 8 7 6 5 Seasonal Swing 2 1.5 1 0.5 0 -0.5 -1 0.5 Residuals 0 -0.5 1985 1990 1995 2000 Time 30 2005 2010 2015 Multiplicative seasonal effects 11 Seasonal plot for All.Beverages 2010 2009 10 2008 1.2 2014 2012 2013 2006 2011 2007 2005 9 2002 2004 2003 8 2010 2013 1.1 1999 2000 1990 1997 2001 1985 1986 1992 1987 1989 1991 1998 1988 1994 1993 1995 1996 7 2011 2015 2014 2008 2012 1.0 1987 1986 2002 2003 1994 1996 1990 1995 1991 1989 1988 1992 1999 2001 1998 2000 0.9 6 2007 2009 2004 2005 2006 1993 1997 Jan - Mar Apr - Jun Jul - Sep Oct - Dec Jan - Mar Quarter Apr - Jun Jul - Sep Quarter 31 Oct - Dec Total volume of beer available for consumption: The overall trend for the total volume of beer available for consumption, shown in the time series plot and the Decomposition graph, show a gradual decrease from 3.8 million litres to 3 million litres. The trend line gradually decreases from 1985 to 1997. It then remains relatively constant, with a slight increase around 2008. From 2008 a relatively moderate decrease can be seen before remaining relatively constant. This decrease may be due to people choosing other types of beverages, as over the years many more options have become available. Looking at the seasonal pattern of the Decomposition graph, the amplitude seems to be slightly decreasing, suggesting a decrease in the volume of beer available for consumption in New Zealand since 1985. The variation of the seasonal pattern ranges between 0 and 1. The seasonal component displayed in the Multiplicative seasonal effects graph shows a high peak in quarter 4 (October- December), followed by a steep decrease in the following quarter (1). The component then decreases further in quarter 2 and then reaching its minimum in quarter 3. The component then shoots back up to its peak from quarter 3 to 4. The high volume of beer available for consumption in quarter 4 is most likely due to the Christmas and New Year holiday, as many people drink more at this time. In addition, because of the summer season, beer is considered a nice cool beverage on hot days, however this should be further investigated. The residuals in the Decomposition graph show no particular pattern. Overall, the residuals have varied between 0.2 and -0.6 since 1985. It is small compared to the trend, and the seasonal component. In addition there is an obvious minimum around 1987, which should be further investigated. Time series plot for Beer 4.5 4 3.5 3 2.5 1985 1990 1995 2000 Time 32 2005 2010 2015 Decomposition of data:Beer Trend Raw data 5 4.5 4 3.5 3 2.5 Seasonal Swing 1 0.5 0 Residuals 0.2 0 -0.2 -0.4 -0.6 1985 1990 1995 2000 Time 33 2005 2010 2015 Multiplicative seasonal effects 1.25 Seasonal plot for Beer 1985 1990 1988 1987 1.20 4.5 1991 1992 1993 1989 1986 1.15 1994 1995 1999 2007 2006 4.0 1986 1989 1.10 2008 1996 1997 2002 2003 2005 2009 2000 2001 2013 1998 2010 2004 2012 2014 2011 1987 1.05 0.95 1.00 1994 1991 2008 1992 1996 1995 2004 2010 1993 2005 2011 2007 2009 1998 1999 2002 2013 2003 2001 2006 1997 2015 2000 2014 2012 0.90 3.0 3.5 1988 1990 Jan - Mar Apr - Jun Jul - Sep Oct - Dec Jan - Mar Quarter Apr - Jun Jul - Sep Quarter 34 Oct - Dec Total volume of Spirits available for consumption: The overall trend for the total volume of spirits available for consumption, shown in the time series plot and the Decomposition graph, shows an increase from 1.2 million litres in 1985 to 2 million litres in 2015. Looking more closely the trend shows a slight peak in 1987 followed by a relative decrease in 1988. There is then a slight increase to 1990, followed by a decrease. From 1998 through till 2011 there is a gradual rise, followed by a relatively gradual decrease. According to stuff.co.nz, the total volume of spirits available has rose because of the increase in spirits stronger than 23 per cent alcohol. With young people being the likely consumers because of the recent binge drinking culture, they are purchasing stronger alcoholic beverages so they can exceed their alcoholic tolerances faster. Looking at the seasonal pattern in the Decomposition graph, the amplitude seems relatively constant until 2005. From 2005 onwards the amplitude seems to increase, suggesting that the data has taken on a more seasonal component in this period, than it had in the past. The variation of the seasonal pattern ranges between 0.8 and -0.4. The seasonal component displayed in the Multiplicative seasonal effects graph shows the maximum peak in quarter 4 (October- December), followed by a steep decrease in the following quarter (1) reaching its minimum. The component then increases significantly from quarter 1 to 2. This could be due to mid-winter festivities such as mid-winter Christmas, winter festivals etc. There is then a slight decrease from quarter 2 to 3, then a significant increase from quarter 3 to 4. Like beer, the large volumes of spirits available for consumption in quarter 4 is most likely due to the Christmas and New Year Holiday. The low volume of spirits in quarter 1 could be due to people making New Year’s resolutions to stop drinking or to stop spending money on alcohol. The residuals in the Decomposition graph show no particular pattern. Overall, the residuals have varied between 0.6 and -0.4 since 1985. It is small compared to the trend and the seasonal component. There are two spikes seen around 1987 and 2009, as well as an obvious minimum around 2012 which should be further investigated. Time series plot for Spirits 3 2.5 2 1.5 1 1985 1990 1995 2000 Time 35 2005 2010 2015 Decomposition of data:Spirits Trend Raw data 3 2.5 2 1.5 1 0.5 0.8 Seasonal Swing 0.6 0.4 0.2 0 -0.2 -0.4 0.6 Residuals 0.4 0.2 0 -0.2 -0.4 1985 1990 1995 2000 Time 36 2005 2010 2015 Seasonal plot for Spirits Multiplicative seasonal effects 1.2 3.0 2009 2010 2012 2013 2.5 2011 2014 2008 2013 2012 2011 2015 2010 2004 2014 1997 2000 2001 1989 1.5 2008 1.1 1999 1998 1992 1990 2009 2004 2005 2006 2007 1988 1987 1985 1991 1993 1994 1995 1996 2002 2003 2001 2000 1.0 1.0 2002 1986 2003 0.9 2.0 2007 2005 2006 1998 1994 1995 1991 1999 1990 1986 1988 1992 1996 1987 1989 1993 1997 Jan - Mar Apr - Jun Jul - Sep Oct - Dec Jan - Mar Quarter Apr - Jun Jul - Sep Quarter 37 Oct - Dec COMPARISON OF THE TWO SERIES The actual recorded volumes of beer available for consumption are greater than spirits. For example the highest volume recorded for beer available for consumption was approximately 4.8 million litres in 1985, whereas the highest volume recorded for spirits available for consumption was 3 million litres in 2009. The two series display distinct patterns in the trend. The amount of beer available for consumption has been gradually decreasing while the amount of spirits has been increasing. More alcoholic beverage options are now available compared to the options available in the 1980’s, which may explain the reduction of beer available for consumption. The increase in the amount of spirits available for consumption could be due to the introduction of ready-made spirit beverages, such as RTD’s, however further investigation should be carried out. Both series show dissimilar seasonality. For beer available for consumption, it is clear that the data is partly influenced by seasonality, as the amplitude of the seasonal pattern in the Decomposition graph seems relatively constant with a slight decrease. For spirits available for consumption, it is clear that the seasonality remains relatively regular, with little variation until 2005. From 2005 to 2015 it is evident that the amplitude increases dramatically, suggesting that the data has been influenced more by seasonality. Both series show relatively similar seasonal patterns when looking at the Multiplicative seasonal effects. Both series shows large volumes available in quarter 4. However the seasonal component for the volume of spirit available for consumption shows relatively higher volumes in quarters 2 and 3 compared to beer. This is probably due to more demand for beer in the summer months, as it is renowned as a nice cool beverage on hot summer days. The variation of the irregular component for the total volume of beer and spirits available for consumption are both smaller than the seasonal component, suggesting that the data is more defined by seasonality. This case seems to be the same for the total volume of all beverages available for consumption data. ACCESSING THE DATA How do I access the Alcohol availability Statistics? This series is found in the Alcohol availability section of the Statistics New Zealand website: www.stats.govt.nz (Statistics NZ Home > Browse for statistics > Industry sectors > Alcohol availability > Alcohol Available for Consumption) How do I get this particular data set and upload it to iNZight? On the Statistics New Zealand website you will find Quick links and Infoshare (www.stats.govt.nz/infoshare). Use the Browse tab. Select Industry Sector> Alcohol Available for Consumption - ALC > Litres of Alcohol (Qrtly-Mar/Jun/Sep/Dec). Under beverage type, hold Ctrl and select: Total beer Total spirit Total all beverage 38 Select all the Time periods. At the bottom right of the screen, change Table on screen to csv file. Read this csv file into iNZight. You will need to remove unnecessary information from the csv first and rename the variables so that only the first row of the csv contains variable names. For this data either imputate the missing values at the beginning of the dataset (i.e. From 1983-1984) or discard these rows. For the example above, the missing values have been discarded. Methods for imputation can be found on: http://www.stats.govt.nz/Census/2013-census/info-about-2013-census-data/2013-census-definitions-forms/definitions/i.aspx Useful links: Are we drinking less beer?-article by Joanna Mathers- http://www.nzherald.co.nz/lifestyle/news/article.cfm?c_id=6&objectid=10871672 Youth drinking cheap spirits a 'real problem' – article by SHANE COWLISHAW http://www.stuff.co.nz/national/health/4685321/Youth-drinking-cheap-spirits-areal-problem Spirits takes up more shelf space for tenth consecutive year- http://www.newswire.co.nz/2013/03/spirits-takes-up-more-shelf-space-for-tenth-consecutive-year/ How to produce Time Series graph, Decomposition graph and Seasonal plot in iNZight? After you have imported your Time Series data click “Advanced” and then “Time Series…” The following command window will appear: 39 Click if your dataset is Additive or Multiplicative. Then select the Time Series Plot option to produce the TOTAL graph, which includes the trend line. Then decompose the dataset using the Decompose option. This will produce the “Decomposition Total” graph. Then select the Seasonplot option to produce the Seasonal plot and Multiplicative/Additive Seasonal Effects. How do you format your time variable so iNZight recognises it? INZight assumes that datasets are quarterly when producing a time series. If your time variable is not in the format the program recognises or if you have not included a quarter column, then you can supply the time structure-information through the module's command window. 1. Click Provide Time Information and then press Create. The following pop-up box will appear: 2 .In Start Date enter the year in which the series started 3. In Frequency enter the number of seasons in a year (e.g. 52 for weekly data) 4. In Season Number enter the number of the season at which the series started (e.g. 3 for 3rd week of the year) 40 Reinterpret the above if your data runs over weeks not years and the seasons are days of the week, or your data concerns days and hours in the day. Why does iNZight not recognise missing values? In regards to data entry, missing values should be coded as NA for iNZight to read the file correctly. However, the time series module will not be able to plot any variables that contain 1 or more missing values. 41 4. Deaths by regional councils of the total population- Multiplicative model Total number of deaths in New Zealand: Time series plot for Total.number.of.deaths.in.New.Zealand 8500 8000 7500 7000 6500 6000 1990 1995 2000 2005 Time 42 2010 2015 Decomposition of data:Total.number.of.deaths.in.New.Zealand 9000 Trend Raw data 8500 8000 7500 7000 6500 6000 5500 Seasonal Swing 1000 500 0 -500 Residuals 400 200 0 -200 -400 1990 1995 2000 Time 43 2005 2010 2015 Multiplicative seasonal effects 8000 1.10 8500 Seasonal plot for Total.number.of.deaths.in.New.Zealand 2012 2011 2009 2013 2004 7000 2011 2008 2007 2010 1.00 2001 2003 2002 1997 1995 2006 2005 2000 1999 2006 2008 2009 1999 2002 1995 2010 1996 2004 2000 1993 1992 2001 2005 2003 1996 1998 1991 1993 1992 1994 1991 1994 1997 0.95 6500 2012 2007 2014 2013 6000 1.05 7500 2014 1998 Jan - Mar Apr - Jun Jul - Sep Oct - Dec Jan - Mar Quarter Apr - Jun Jul - Sep Quarter 44 Oct - Dec Number of deaths – Northland region The trend for the number of deaths in the Northland region, illustrated on the time series plot, shows an increase from 260 in 1991, to 350 in 2015. However the trend also shows dip, evidently in 1996, 1998, 2001, between 2005-2008 and 2011. In 2004 and 2009 two distinct peaks are evident. The number of deaths has gradually increased over time possibly due to population growth, particularly in the older ages, partly offset by longer life expectancy. The detail of the seasonal pattern can be read in the Multiplicative Seasonal Effects graph, where the number of deaths are lowest in the first quarter (Jan-Mar) and highest in quarter three (July-Sep). There is a slight increase from quarter 1 to 2, then a steep increase from quarter 2 to 3, and then the seasonal component shows a relatively steep decrease from quarter 3 to 4. The seasonal component then decreases further from quarter 4 to 1. This may be due to the cooler weather around this time of year resulting in road accidents due to ice, fog, and frost as well as heightened flu symptoms, viruses, etc. Looking at the seasonal pattern in the Decomposition graph, the amplitude seems relatively constant from 1991 to 2015 in the number of deaths in the Northland Region, varying from 40 to -20. In addition, the residuals in the Decomposition shows that the irregular has no particular pattern, and has varied between 30 and -30 since 1991. It is small compared to the trend and relatively similar to the seasonal component. Time series plot for Northland.Region 350 300 250 1990 1995 2000 2005 Time 45 2010 2015 Decomposition of data:Northland.Region Trend Raw data 400 350 300 250 200 Seasonal Swing 40 30 20 10 0 -10 -20 Residuals 30 20 10 0 -10 -20 -30 1990 1995 2000 Time 46 2005 2010 2015 Seasonal plot for Northland.Region Multiplicative seasonal effects 350 1.10 2011 2014 2009 2013 2012 2008 2014 2002 2007 2003 2010 2004 2007 2003 2009 1999 2002 2010 2013 1999 1995 1997 2001 2000 1998 2006 2005 1996 1994 2000 2001 1.00 2004 2006 2005 1996 1995 1997 1993 1993 1991 1992 1998 1994 0.95 300 2011 2008 250 1.05 2012 1992 1991 Jan - Mar Apr - Jun Jul - Sep Oct - Dec Jan - Mar Quarter Apr - Jun Jul - Sep Quarter 47 Oct - Dec Number of Deaths – Southland region The trend for the number of deaths in the Southland region, illustrated on the time series plot, has remained relatively constant, around 210, from 1991 to 2015. The trend shows a slight decrease of 30 from 1991 to 2006, then rises back up to 210 by 2015. The detail of the seasonal pattern can be read in the “Multiplicative Seasonal Effects” graph, where the number of deaths are lowest in the first quarter (Jan-Mar), highest in quarter 3 (July-Sep) and relatively the same in quarters 2 and 4. This may be due to the cooler weather around July-September, resulting in road accidents due to ice, fog, and frost as well as heightened flu symptoms, viruses, etc. There is a slight increase from quarter 1 to 2, then a steep increase from quarter 2 to 3, and then the seasonal component shows a relatively steep decrease from quarter 3 to 4. The seasonal component then decreases further, quite steeply, from quarter 4 to 1. Looking at the seasonal pattern in the Decomposition graph, the amplitude seems relatively constant from 1991 to 2015 in the number of deaths in the Southland Region, varying from 20 to -10. In addition, the residuals in the Decomposition shows that the irregular has no particular pattern, and has varied between 30 and -20 since 1991. It is small compared to the trend, however it is larger than the seasonal component. Time series plot for Southland.Region 260 240 220 200 180 160 1990 1995 2000 2005 Time 48 2010 2015 Decomposition of data:Southland.Region 280 Trend Raw data 260 240 220 200 180 160 140 Seasonal Swing 20 10 0 -10 Residuals 30 20 10 0 -10 -20 1990 1995 2000 Time 49 2005 2010 2015 Multiplicative seasonal effects 240 1.10 260 Seasonal plot for Southland.Region 220 2014 2002 2007 1994 1998 2008 1991 2006 1991 1992 2000 2014 1994 2011 1999 2001 2003 2009 2004 2011 1992 2000 2001 1997 1996 1.00 200 1.05 1995 2012 1993 2013 2007 1999 2006 1995 1996 2013 2002 2003 2004 0.95 180 2010 160 2005 2012 1993 1997 1998 2010 2005 2009 2008 Jan - Mar Apr - Jun Jul - Sep Oct - Dec Jan - Mar Quarter Apr - Jun Jul - Sep Quarter 50 Oct - Dec COMPARISON OF THE TWO SERIES: The number of deaths in the Northland region are marginally higher than those in the Southland region. The two series display distinct patterns in the trend, with the number of deaths in the Northland region showing an increase compared to the Southland region, remaining relatively constant. This is most likely due to population growth, particularly in the older ages, in the North Island. Older people may choose to live in regions such as Northland as it is warmer and housing prices are cheaper, hence more numbers. Both series as well as the total series show relatively similar seasonal patterns. All of the seasonal patterns show high numbers of deaths in quarter 3 and low numbers in quarter 1. This suggests that a national reason, such as the winter season or sickness may be related to the high numbers of death around this time of year. The irregulars for the number of deaths in the Northland and Southland region show no particular pattern. ACCESSING THE DATA How do I access the Population Statistics? This series is found in the Population section of the Statistics New Zealand website: www.stats.govt.nz (Statistics NZ Home > Browse for statistics > Population > Deaths) How do I get this particular data set and upload it to iNZight? On the Statistics New Zealand website you will find Quick links and Infoshare (www.stats.govt.nz/infoshare). Use the Browse tab. Select Population> Deaths- VSD > Deaths by area, regional councils (Maori and total population) (Qrtly-Mar/Jun/Sep/Dec). Under Regional Council, hold Ctrl and select: Northland Region Southland Region Total New Zealand Under Ethnic group, select “Total” Select all the Time periods. At the bottom right of the screen, change Table on screen to csv file. Read this csv file into iNZight. You will need to remove unnecessary information from the csv first and rename the variables so that only the first row of the csv contains variable names. 51 Useful links: Death Statistics- http://www.stats.govt.nz/browse_for_stats/population/deaths.aspx How to produce Time Series graph, Decomposition graph and Seasonal plot in iNZight? After you have imported your Time Series data click “Advanced” and then “Time Series…” The following command window will appear: Click if your dataset is Additive or Multiplicative. Then select the Time Series Plot option to produce the TOTAL graph, which includes the trend line. Then decompose the dataset using the Decompose option. This will produce the “Decomposition Total” graph. Then select the Seasonplot option to produce the Seasonal plot and Multiplicative/Additive Seasonal Effects. How do you format your time variable so iNZight recognises it? INZight assumes that datasets are quarterly when producing a time series. If your time variable is not in the format the program recognises or if you have not included a quarter column, then you can supply the time structure-information through the module's command window. 1. Click Provide Time Information and then press Create. 52 The following pop-up box will appear: 2 .In Start Date enter the year in which the series started 3. In Frequency enter the number of seasons in a year (e.g. 52 for weekly data) 4. In Season Number enter the number of the season at which the series started (e.g. 3 for 3rd week of the year) Reinterpret the above if your data runs over weeks not years and the seasons are days of the week, or your data concerns days and hours in the day. 53 5. Number of pigs killed in New Zealand- Multiplicative model Total number of pigs killed in New Zealand: Time series plot for Total.Pigs.killed.in.New.Zealand 80000 75000 70000 65000 60000 55000 50000 1990 1995 2000 2005 Time 54 2010 2015 Decomposition of data:Total.Pigs.killed.in.New.Zealand Trend Raw data 80000 75000 70000 65000 60000 55000 50000 45000 Seasonal Swing 4000 2000 0 -2000 -4000 6000 Residuals 4000 2000 0 -2000 -4000 -6000 1990 1995 2000 Time 55 2005 2010 2015 Seasonal plot for Total.Pigs.killed.in.New.Zealand Multiplicative seasonal effects 75000 1992 1.05 1993 70000 1998 1997 2003 1991 1995 1994 1988 1996 2004 2005 1999 2008 2002 2007 1989 1990 2010 2008 1995 50000 2003 2004 1992 1997 2011 2006 2009 2001 2000 2013 2012 2014 1991 2006 1990 1998 2001 1994 2012 2005 2013 1993 1989 2002 1999 2011 2009 2014 2000 2010 0.95 55000 60000 2007 1.00 65000 1996 2015 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Month Feb Mar Apr May Jun Jul Month 56 Aug Sep Oct Nov Dec Number of pigs killed– North Island The trend for the number of pigs killed in the North Island regions, illustrated in the time series plot and the Decomposition graph, shows a gradual decrease from 39000 in 1988, to 21000 in 2015. A total decrease of 18000 pig kills in the course of 27 years. This decrease in pig kills could be due to the fact that production costs for pig farmers have increased drastically, while prices for pork have increased only marginally in the same period. Other challenges faced by the pig industry may include competition from imported pork, animal disease outbreaks, and increasing costs associated with animal welfare. The seasonal pattern, illustrated in the Decomposition graph, has remained relatively regular, however, the magnitude decreased slightly around 1997. There is a peak in December, followed by a decrease in January. This might be due to the Christmas Holiday, as many people consume ham only at this time of year. There is another smaller peak in March (possibly due to the Easter/school holidays). Next, there is a slight decrease in April, however the number of kills increase again in May. There is a steep decrease in June, until another rise in July, remaining constant till August, (which may be due to school holidays), then again decreasing in September. From October onwards the trend begins to climb steeply again until it reaches its global maximum in December. In addition, the residuals in the Decomposition shows that the irregular has no particular pattern, and has varied between 3000 and -3000 since 1988. It is small compared to the trend and relatively similar to the seasonal component. Time series plot for North.Island.Regions 45000 40000 35000 30000 25000 20000 1990 1995 2000 2005 Time 57 2010 2015 Decomposition of data:North.Island.Regions Trend Raw data 45000 40000 35000 30000 25000 20000 Seasonal Swing 3000 2000 1000 0 -1000 -2000 -3000 3000 Residuals 2000 1000 0 -1000 -2000 -3000 1990 1995 2000 Time 58 2005 2010 2015 Multiplicative seasonal effects 45000 Seasonal plot for North.Island.Regions 1.05 40000 1988 1992 1989 1991 1993 1998 1995 1997 35000 1994 1996 1990 1999 2003 1993 1997 1998 1994 2001 1999 2002 2002 2000 2001 2003 2010 2009 2006 2005 2004 2008 2007 2004 2008 2000 0.95 25000 30000 1992 1995 1996 20000 1.00 1990 1989 1991 2011 2007 2012 2014 2013 2013 2006 2009 2005 2012 2014 2011 2010 2015 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Month Feb Mar Apr May Jun Jul Month 59 Aug Sep Oct Nov Dec Number of pigs killed – South Island The trend for the number of pigs killed in the South Island regions, illustrated in the time series plot and the Decomposition graph, shows a significant increase from 1988 to 1996, followed by a clear decrease from 1996 to around 2002. This is then followed by the steep increase to 2008, followed by a steady decrease of 18.4% from 2008 to 2015. The increase in pig kills could be due to the fact that because Canterbury holds the largest pig-farming industry in the country and is a large grain-growing region, pig farmers had easier access to feed pigs, allowing them to become large enough for human consumption and slaughtered. Also, an increase in the number of entrants in hunting competitions in the South Island, such as the annual Highwayman Hotel Pig Hunting Competition at Dunback, could explain the reason for the increase in the number of kills, setting record numbers in 2009 according to the Otago Daily Times website. This may explain the peak numbers in 2009 on the graph. The decrease from 2008-2015 is most likely due to competition for imported pork. The magnitude of the seasonal pattern in the Decomposition has remained relatively constant between 2000 and -2000 since 1988. The Multiplicative seasonal effects graph shows the highest peak in May, and lowest numbers in February and September. Further investigation should be conducted to find reason for these peaks and falls during these months. In addition, the residuals in the Decomposition shows that the irregular has no particular pattern, and has varied between 2000 and -4000 since 1988. It is small compared to the trend but large compared to the seasonal component. Time series plot for South.Island.Regions 40000 35000 30000 25000 1990 1995 2000 2005 Time 60 2010 2015 Decomposition of data:South.Island.Regions Trend Raw data 45000 40000 35000 30000 25000 20000 3000 Seasonal Swing 2000 1000 0 -1000 -2000 Residuals 2000 0 -2000 -4000 1990 1995 2000 Time 61 2005 2010 2015 Multiplicative seasonal effects 1.08 Seasonal plot for South.Island.Regions 40000 2003 2004 2007 2008 1993 1992 2010 2011 2012 2004 1996 2006 2011 2003 2005 1997 2001 1.02 2006 2009 2002 1997 2013 1994 1999 1998 2012 2014 1995 1996 1991 2014 2013 2009 2010 1995 1.00 35000 1.04 2005 2007 2001 1994 2000 1998 2015 2002 1999 1992 2000 0.98 30000 1.06 2008 1993 0.96 1990 1988 1990 1989 1989 Jan 0.94 25000 1991 Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Month Feb Mar Apr May Jun Jul Month 62 Aug Sep Oct Nov Dec COMPARISON OF THE TWO SERIES: The two series display distinct patterns in the trend, with the number of pig kills in North Island regions showing a gradual decrease of 46.2% compared to the South Island regions, showing a 34.2% increase from 1988 to 2008, then a steady decrease of 18.4% from 2008 to 2015. The South Island holds the largest pig-farming industry in the country and is a large grain-growing region, therefore, pig farmers have easier access to feed pigs, allowing them to become large enough for human consumption. Also the increase in the number of entrants for hunting competitions in the South Island, such as the annual Highwayman Hotel Pig Hunting Competition at Dunback, could explain the reason for the increase in the number of kills. However, like the North Island, the effects of competition with imported pork as well as increasing costs associated with animal welfare could explain the decreases. The Multiplicative seasonal effects show dissimilar peaks and troughs between the North and South Island regions. The North Island region shows the largest peak in December while the South Island shows the largest peak in May. The peak in December could be explained by the Christmas holiday, as many families choose to consume ham at this time of year, therefore increasing need in markets. As for the peak in May in the South Island, further investigation should be conducted to find reason for this. The irregulars for the number of pig kills in the North Island and South Island regions show no particular pattern. Both series as well as the total series show reasonably regular seasonal patterns since 1988. However it should be noted that the variation of the seasonal pattern for both the total and South Island data is smaller than the irregular, which indicates that these time series under analysis is dominated by its irregular component, limiting accurate predictions or forecasts. In contrast, the variation of the seasonal pattern for the North Island is exactly the same as the variation of the residuals, indicating that this time series under analysis is neither dominated by the irregular or seasonal, therefore, caution should be taking when formulating any forecasts or predictions. ACCESSING THE DATA How do I access the Livestock slaughtering Statistics? This series is found in the Agriculture, horticulture and forestry section of the Statistics New Zealand website: www.stats.govt.nz (Statistics NZ Home > Browse for statistics > Industry sectors > Agriculture, horticulture, and forestry) How do I get this particular data set and upload it to iNZight? On the Statistics New Zealand website you will find Quick links and Infoshare (www.stats.govt.nz/infoshare). Use the Browse tab. Select Population> Deaths- VSD > Deaths by area, regional councils (Maori and total population) (Qrtly-Mar/Jun/Sep/Dec). Under Regional Council, hold Ctrl and select: Northland Region Southland Region Total New Zealand Under Ethnic group, select “Total” 63 Select all the Time periods. At the bottom right of the screen, change Table on screen to csv file. Read this csv file into iNZight. You will need to remove unnecessary information from the csv first and rename the variables so that only the first row of the csv contains variable names. Useful links: Otago Daily Times, “Record set pig kills”- http://www.odt.co.nz/the-regions/north-otago/66012/records-set-pig-kill Agriculture Statistics- http://www.stats.govt.nz/browse_for_stats/industry_sectors/agriculture-horticulture-forestry.aspx How to produce Time Series graph, Decomposition graph and Seasonal plot in iNZight? After you have imported your Time Series data click “Advanced” and then “Time Series…” The following command window will appear: Click if your dataset is Additive or Multiplicative. Then select the Time Series Plot option to produce the TOTAL graph, which includes the trend line. Then decompose the dataset using the Decompose option. This will produce the “Decomposition Total” graph. Then select the Seasonplot option to produce the Seasonal plot and Multiplicative/Additive Seasonal Effects. 64 How do you format your time variable so iNZight recognises it? INZight assumes that datasets are quarterly when producing a time series. If your time variable is not in the format the program recognises or if you have not included a quarter column, then you can supply the time structure-information through the module's command window. 1 .Click Provide Time Information and then press Create. The following pop-up box will appear: 2. In Start Date enter the year in which the series started 3 .In Frequency enter the number of seasons in a year (e.g. 52 for weekly data) 4. In Season Number enter the number of the season at which the series started (e.g. 3 for 3rd week of the year) Reinterpret the above if your data runs over weeks not years and the seasons are days of the week, or your data concerns days and hours in the day 65 6. Gross weight (tonnes) of imports by seaports in New Zealand- Multiplicative model Total gross weight of imports by seaports in New Zealand: Time series plot for Total.Gross.Weight.of.imports.by.Seaports.in.NZ 2e+06 1500000 1e+06 5e+05 1990 1995 2000 2005 Time 66 2010 2015 Decomposition of data:Total.Gross.Weight.of.imports.by.Seaports.in.NZ 2500000 Trend Raw data 2e+06 1500000 1e+06 5e+05 150000 1e+05 50000 0 -50000 -1e+05 -150000 -2e+05 Seasonal Swing Residuals 4e+05 2e+05 0 -2e+05 -4e+05 1990 1995 2000 Time 67 2005 2010 2015 2006 2014 2004 2013 2014 2007 2012 2009 2005 2002 2000 1999 2010 2005 1997 2001 2001 1999 1998 1994 1997 1996 2000 1990 1998 1988 1995 1994 1991 1995 1996 1991 1992 1993 1992 1990 1989 1993 1988 1989 Jan 1.00 2007 2010 2008 2006 2003 2012 2011 2002 2015 2008 2011 2004 2009 0.95 1000000 1.05 2013 2003 500000 Multiplicative seasonal effects Feb Mar Apr May Jun Jul Aug Sep Oct Nov 0.90 1500000 2000000 Seasonal plot for Total.Gross.Weight.of.imports.by.Seaports.in.NZ Dec Jan Month Feb Mar Apr May Jun Jul Month Gross weight of imports- Auckland port: 68 Aug Sep Oct Nov Dec The trend for the gross weight (tonnes) of imports for the Auckland seaport, illustrated in the time series plot and the Decomposition graph, reveals a gradual increase from 140,000 tonnes in 1988, to 390,000 tonnes in 2015. A total increase of 250,000 tonnes over the course of 27 years. Two dips are evident in the trend. One in 2009 and one around 2012/13. This increase in gross weight of imports resembles closely to the total data and could be due to increases in mineral fuels as mineral fuels are the largest component of sea-freighted cargo imports by both weight and value in New Zealand. It could also be due to increases in cargo weight by legislation, as a result of increasing demand. According to a report called the the “30 50 Port of Whangarei”, the dip in 2009 was mainly due to decreases in mineral fuels, mineral oils and products. In addition, the dip in 2012/13 is probably also due to decreases in mineral fuels, however further investigation is needed. Looking at the seasonal pattern of the Decomposition graph, the amplitude seems to be increasing, suggesting an increase in gross weight (tonnes) of imports into New Zealand since 1988. The variation of the seasonal pattern ranges between -20000 and 40000. The seasonal component displayed in the Multiplicative seasonal effects graph shows a peak in November, followed by steep decrease in December. This might be due to the Christmas Holiday, as the demand during the lead up to Christmas for imports such as food, toys, etc increases. From December, import gross weight decreases, reaching its lowest gross weight in February. From February through till May, there is a gradual increase in import gross weight, most likely due to preparation for the winter season, as things such as timber, oil, heating electronics etc may be of more demand. From May, the seasonal component shows a slight decrease in June, and from June to November the seasonal component increases radically. The seasonal plot shows a record high gross weight of imports in April 2014. Further investigation should be carried out to identify the cause of this spike in gross weight. In addition, if we exclude the maximum residuals in 1994 and 1998, the residuals in the Decomposition graph show an increase in the variation from 2005 onwards. This suggests that the residuals have become more volatile and could possibly be due to the quality of the model. Overall, the residuals have varied between 200,000 and -100,000 since 1988. It is small compared to the trend, however it is large compared to the seasonal component. Time series plot for Auckland.Seaport 5e+05 4e+05 3e+05 2e+05 1e+05 1990 1995 2000 2005 Time 69 2010 2015 Decomposition of data:Auckland.Seaport Trend Raw data 5e+05 4e+05 3e+05 2e+05 1e+05 Seasonal Swing 40000 20000 0 -20000 2e+05 Residuals 150000 1e+05 50000 0 -50000 -1e+05 1990 1995 2000 Time 70 2005 2010 2015 Multiplicative seasonal effects 4e+05 1.10 5e+05 Seasonal plot for Auckland.Seaport 2014 2009 2013 1.05 2004 2003 2008 2009 2006 2002 2011 2005 1997 2012 2011 1996 1994 1998 2000 1995 2e+05 2012 2010 1999 1997 1996 1998 2001 1e+05 1.00 2010 2003 2007 2002 2004 2006 1999 2005 2008 2001 1992 1990 2000 1991 1995 1988 1994 1993 1989 1991 1989 1992 1993 0.95 3e+05 2014 2007 2013 2015 1988 1990 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Month Feb Mar Apr May Jun Jul Month 71 Aug Sep Oct Nov Dec Gross weight of imports- Tauranga port: The trend for the gross weight (tonnes) of imports for the Tauranga seaport, illustrated in the time series plot and the Decomposition graph, reveals an increase from 90,000 tonnes in 1988, to 390,000 tonnes in 2015. A total increase of 300,000 tonnes over the course of 27 years. From 1988 to 1998, a very steady increase in gross weight of imports can be seen. Then from 1999 to 2006, it is noticeable that the gross weight of imports increases much faster. In 2007 a decrease is evident, however, in 2008, another increase occurs, followed by another in 2009. From 2009 to 2014 the gross weight of imports has been gradually increasing. A slight decrease in 2015 is also evident. Looking at the seasonal pattern of the Decomposition graph, the amplitude seems to have increased drastically from 2004 to 2015, suggesting that the data has taken on a more seasonal component in this period, than it had in the past. The variation of the seasonal pattern ranges between -80000 and 40000. The seasonal component displayed in the Multiplicative seasonal effects graph shows the largest peak in July and the lowest trough in December. Other peaks are noticeable in March, October and November. There is a steep increase from January through until March, followed by a regularly steep decrease from March until June. There is a massive rise in gross weight of imports from June to July, followed by a slight decrease in August, then a steady rise to October, remaining relatively constant until November. From November to December a very steep decrease is evident. The seasonal plot shows a record high gross weight of imports in September 2014. This could be due to trade volumes of the port of Tauranga increasing by 3.5% to 19,736,902 tonnes or the acquirement of 50% shareholding in PrimePort Timaru Limited. However further investigation should be carried out to answer this. In addition, the residuals in the Decomposition graph shows that the irregular has no particular pattern. Because of the enormous peak in 2014, the variation of the residuals is large, ranging from 500,000 and -100,000 since 1988. It is small compared to the trend, however it is large compared to the seasonal component, suggesting that the data is not well defined by seasonality. Time series plot for Tauranga.Seaport 8e+05 6e+05 4e+05 2e+05 0 1990 1995 2000 2005 Time 72 2010 2015 Decomposition of data:Tauranga.Seaport 1e+06 Trend Raw data 8e+05 6e+05 4e+05 2e+05 0 Seasonal Swing 5e+05 40000 20000 0 -20000 -40000 -60000 -80000 Residuals 4e+05 3e+05 2e+05 1e+05 0 -1e+05 1990 1995 2000 Time 73 2005 2010 2015 Multiplicative seasonal effects 2014 2006 2013 2009 0.90 2012 2005 2004 2008 2007 2011 2007 2013 2012 2008 2011 2010 2014 2006 2015 2009 2005 2004 2003 2010 0.85 2e+05 4e+05 0.95 1.00 6e+05 1.05 8e+05 1.10 Seasonal plot for Tauranga.Seaport 2002 0e+00 2002 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov 0.80 1995 2001 1999 2000 1997 1992 1994 1998 1989 1996 1991 1988 1993 1990 2003 1997 1999 1991 2000 1995 1998 1996 1988 2001 1993 1994 1990 1989 1992 Dec Jan Month Feb Mar Apr May Jun Jul Month 74 Aug Sep Oct Nov Dec COMPARISON OF THE TWO SERIES: It is clear in both series that the gross weight of imported cargo through continues to increase. The major mode of transport for imported cargo is by sea as over 99% of all imported cargo arrives by shipping. According to the Statistics New Zealand webpage, Whangarei, Auckland and Tauranga combined handled over two-thirds of the gross weight of sea imports that arrived into New Zealand for the 12 month period ending June 2013 (Figure 2). The actual recorded gross weight of imports for the Tauranga seaport are higher than the Auckland seaport. For example the highest gross weight recorded for Tauranga seaport is around 900,000 tonnes whereas the highest gross weight recorded for Auckland seaport is approximately 500,000 tonnes. However, this is due to a single massive spike recorded around 2014 for the Tauranga port. With spike removed, the actual recorded gross weight of imports are relatively the similar. The two series display relatively similar patterns in the trend, showing increases. However, the trend for the gross weight of imports for the Auckland port shows a more significant and steeper increase in the trend than the Tauranga port. Both series show dissimilar seasonality. For the Auckland port, it is clear that over the past 27 years that gross weight of imports is partly due to seasonality, as the amplitude of the seasonal pattern, displayed in the Decomposition graph remains relatively constant. However, it is difficult to validate this as the variation of the seasonal pattern is smaller than the residuals, suggesting that the residuals influence the Auckland port data more. For the Tauranga port, the seasonality remains relatively regular, with little variation until 2004. From 2004 to 2015 it is evident that the amplitude increases dramatically, suggesting that the data has been influenced more by seasonality. Both series show distinct seasonal patterns when looking at the Multiplicative seasonal effects. Auckland port shows maximum gross weight of imports in November, while Tauranga port shows maximum gross weight of imports in July. The variation of the irregular component for the Auckland and Tauranga port are both larger than the seasonal component, suggesting that the data is more defined by the irregular. This case seems to be the same for the Total gross weight of imports to seaports data. ACCESSING THE DATA How do I access the Overseas Cargo Statistics? This series is found in the Imports and exports section of the Statistics New Zealand website: www.stats.govt.nz (Statistics NZ Home > Browse for statistics > Industry sectors > Imports and exports > Overseas Cargo Statistics) How do I get this particular data set and upload it to iNZight? On the Statistics New Zealand website you will find Quick links and Infoshare (www.stats.govt.nz/infoshare). Use the Browse tab. Select Imports and exports> Overseas Cargo Statistics - OSC> Total Imports by New Zealand Port (Monthly) Under New Zealand Port, hold Ctrl and select: Auckland (sea) 75 Tauranga (sea) Total Seaports Under observations, select “Gross weight (tonnes)” Select all the Time periods. At the bottom right of the screen, change Table on screen to csv file. Read this csv file into iNZight. You will need to remove unnecessary information from the csv first and rename the variables so that only the first row of the csv contains variable names. Useful links: http://www.port-tauranga.co.nz/Community/History/ http://www2.stats.govt.nz/domino/external/PASFull/pasfull.nsf/00c9e0a06fee31764c2568470008f782/4c2567ef00247c6acc256da6001934e6?OpenDocument http://www.ehinz.ac.nz/assets/Factsheets/Released-2013/EHI20-21-GrossWeightOfImportedCargoIntoNZ2009-2013-released201310.pdf http://www.wdc.govt.nz/PlansPoliciesandBylaws/Plans/SustainableFutures/Documents/Sustainable%20Infrastructure/30-50-Port-of-Whangarei.pdf How to produce Time Series graph, Decomposition graph and Seasonal plot in iNZight? After you have imported your Time Series data click “Advanced” and then “Time Series…” The following command window will appear: 76 Click if your dataset is Additive or Multiplicative. Then select the Time Series Plot option to produce the TOTAL graph, which includes the trend line. Then decompose the dataset using the Decompose option. This will produce the “Decomposition Total” graph. Then select the Seasonplot option to produce the Seasonal plot and Multiplicative/Additive Seasonal Effects. How do you format your time variable so iNZight recognises it? INZight assumes that datasets are quarterly when producing a time series. If your time variable is not in the format the program recognises or if you have not included a quarter column, then you can supply the time structure-information through the module's command window. 1 .Click Provide Time Information and then press Create. The following pop-up box will appear: 77 2 .In Start Date enter the year in which the series started 3. In Frequency enter the number of seasons in a year (e.g. 52 for weekly data) 4. In Season Number enter the number of the season at which the series started (e.g. 3 for 3rd week of the year) Reinterpret the above if your data runs over weeks not years and the seasons are days of the week, or your data concerns days and hours in the day. 78 7. Persons employed in Labour Force Aged 15-19 and 20-24- Multiplicative model Total all ages employed in Labour Force in New Zealand: Time series plot for Total.for.all.ages.employed.in.LF.in.NZ 2400 2200 2000 1800 1600 1985 1990 1995 2000 2005 Time 79 2010 2015 Decomposition of data:Total.for.all.ages.employed.in.LF.in.NZ Trend Raw data 2400 2200 2000 1800 1600 1400 Seasonal Swing 20 10 0 -10 -20 -30 15 10 5 0 -5 -10 Residuals 1985 1990 1995 2000 2005 Time 80 2010 2015 2015 2014 2014 2013 2012 2011 2007 2009 2010 2008 2008 2011 2007 2010 2012 2009 2006 2006 2005 1.006 2013 1.004 2200 Multiplicative seasonal effects 1.008 Seasonal plot for Total.for.all.ages.employed.in.LF.in.NZ 2004 2003 2003 2002 1600 2000 2001 1999 2000 1999 1998 1997 1996 1997 1996 1998 1995 1995 1994 1987 1986 1987 1986 1988 1994 1993 1988 1990 1989 1992 1991 1989 1990 1993 1991 1992 Jan - Mar 1.000 2001 0.998 1800 2002 1.002 2004 0.996 2000 2005 Apr - Jun Jul - Sep Oct - Dec Jan - Mar Quarter Apr - Jun Jul - Sep Quarter 81 Oct - Dec Labour Force aged 15 to 19: The trend for people employed in Labour Force aged 15 to 19, illustrated in the time series plot and the Decomposition graph, reveal an overall decrease from 190 in 1986, to 120 in 2015. A total decrease of 70 individuals aged 15-19 over the course of 29 years. From 1986 to 1993, there was a steep decrease, followed by a steady increase up to 1996 and remaining constant till 1998. Following there was a slight decrease from 1998 to 1999 until a gradual increase up until 2008. This is then followed by a steep decrease till 2013, then the trend begins to increase sharply again to 2015. Looking at the seasonal pattern of the Decomposition graph, the amplitude seems relatively regular, however, it should be noted that even though the difference is minimal there is a slight decrease in amplitude around 1993 further decreasing in 2008. The variation of the seasonal pattern ranges between -4 and 4. The seasonal component displayed in the Multiplicative seasonal effects graph shows a massive peak in quarter 4 (October- December) and a trough in quarters 2 and 3 (April-September). This might be due to the summer school holidays. Teenagers are always looking for work during the summer and summer is the busiest time of year for retail, tourism and hospitality, therefore employers open up more positions at this time of year. During the winter period, students are too focused on their school work. In addition the residuals in the Decomposition graph show no particular pattern. Overall, the residuals have varied between 4 and - since 1986. It is small compared to the trend, however it is large compared to the seasonal component, meaning that the irregular has some form on influence on the data. Time series plot for Aged.15.19 180 160 140 120 100 1985 1990 1995 2000 2005 Time 82 2010 2015 Decomposition of data:Aged.15.19 Trend Raw data 200 180 160 140 120 100 80 Seasonal Swing 6 4 6 4 2 0 -2 -4 Residuals 2 0 -2 -4 -6 -8 1985 1990 1995 2000 Time 83 2005 2010 2015 Multiplicative seasonal effects 1.04 Seasonal plot for Aged.15.19 1986 1.03 1987 180 1986 1987 160 1.02 1988 1989 2004 2008 2008 2005 2001 2002 1995 2000 1997 1999 2014 1996 1991 2009 1994 1998 1999 2014 1994 2012 1993 2011 1992 1993 2013 2010 2011 0.98 1997 1996 1998 2015 2001 2000 1992 1995 2010 2013 2012 0.97 120 100 1.00 1990 2003 2004 2002 2003 2009 1991 0.99 140 1989 2007 2006 1990 1.01 1988 2007 2006 2005 Jan - Mar Apr - Jun Jul - Sep Oct - Dec Jan - Mar Quarter Apr - Jun Jul - Sep Quarter 84 Oct - Dec Labour Force aged 20 to 24: The trend for people employed in Labour Force aged 20 to 24, illustrated in the time series plot and the Decomposition graph, reveal a decrease of 238 to from 1986 to 167 in 2001. Evidently following there is then an increase back up from 167 to 230 from 2001 to 2015. The dip in 2001 could be due to a national economic crisis whereby businesses were unable to take on young employees/recent graduates, however further investigation should be carried out to validate this. Looking at the seasonal pattern in the Decomposition graph, the amplitude seems relatively regular, suggesting that the seasonal component could be the cause of the number of individuals aged 20 to 24 in the Labour Force since 1986. The variation of the seasonal pattern ranges between 4 and -4. The seasonal component displayed in the Multiplicative seasonal effects graph shows a massive peak in quarter 4 (October- December) and a trough in quarters 2 and 3 (April-September). This might be due to the summer university/polytechnic/institutes holidays. Students are always looking for work during the summer after exams and summer is the busiest time of year for retail, tourism and hospitality, therefore employers open up more positions at this time of year. During the winter period, students are too focused on their education. In addition the residuals in the Decomposition graph show no particular pattern. Overall, the residuals have varied between 6 and -6 since 1986. It is small compared to the trend, however it is large compared to the seasonal component, meaning that the irregular has a larger influence on the data then the seasonal component. Time series plot for Aged.20.24 240 220 200 180 160 1985 1990 1995 2000 2005 Time 85 2010 2015 Decomposition of data:Aged.20.24 Trend Raw data 240 220 200 180 160 Seasonal Swing 4 2 0 -2 8 -4 Residuals 6 4 2 0 -2 -4 -6 1985 1990 1995 2000 Time 86 2005 2010 2015 Multiplicative seasonal effects 1.020 240 Seasonal plot for Aged.20.24 1986 1986 1.015 1987 2014 2015 1.010 220 1987 2014 1988 1.005 2013 2013 1995 2007 2011 2012 1996 1994 1989 1990 180 1992 2003 1998 2002 2002 1999 1999 2001 0.995 2004 2008 1997 1993 2009 1991 2010 2005 1.000 2007 1995 2006 1988 2004 2008 2010 2012 1994 2011 1989 1996 2005 1990 1993 1992 2003 1997 1991 2009 0.990 200 2006 2000 1998 160 0.985 2001 2000 Jan - Mar Apr - Jun Jul - Sep Oct - Dec Jan - Mar Quarter Apr - Jun Jul - Sep Quarter COMPARISON OF THE TWO SERIES: 87 Oct - Dec The actual values of people aged 20-24 in the Labour Force (240) are greater than those people aged 15-19 (190). This is expected as a lot of people aged 20-24 have finished both high school and university and are looking for employment, whereas, people aged 15-19 are most likely still at school or starting university, focusing more on study than work. The trends for both series are quite distinct. People aged between 15 and 19 reveal an overall decrease, whereas, people aged between 20 and 24 show some form of recovery after a massive decrease from 1986 to 2001. The trend shows that in 2015 the number of people in the labour force between 20 and 24 have almost recovered back to its original numbers in 1986. The recovery for the people aged 20-24 year olds could possibly be due to employees wanting more graduates within their business as most graduates display more motivation to learn, have new ideas and are cheaper to employ, benefitting the business. Most people aged 15 to 19 have yet to achieve their qualifications that are well sought for in the Labour Force. The decrease in the number of people aged between 15 to 19 years may be due to employers becoming stricter on qualification levels and age since 1986. The seasonal component of both series displayed in the Multiplicative seasonal effects graphs both show peaks in quarter 4 (October- December) and a trough in quarters 2 and 3 (April-September). These are likely due to the school and university holidays. Teenagers and students are always looking for work during the summer and summer is the busiest time of year for retail, tourism and hospitality, therefore employers open up more positions at this time of year. During the winter period, high school and university students are too focused on their school work. In addition the irregular for both series are larger than the seasonal pattern, suggesting that for both these series, the irregular has a larger influence on the data then the seasonal component. When comparing both these series to the total all ages employed in Labour Force in New Zealand, it is clear that the people aged 15 to 19 are only a small component of this data as the decrease is masked. The data for people aged 20-24 could be influencing the increase in the overall ages of people employed in the Labour Force, however, this group is also small, therefore further investigation should be carried out about all age groups to validate the increase. ACCESSING THE DATA How do I access the Labour Force Statistics? This series is found in the Employment and unemployment section of the Statistics New Zealand website: www.stats.govt.nz (Statistics NZ Home > Browse for statistics > Income and work > Employment and unemployment > Household Labour Force Survey) How do I get this particular data set and upload it to iNZight? On the Statistics New Zealand website you will find Quick links and Infoshare (www.stats.govt.nz/infoshare). Use the Browse tab. Select Work income and spending > Household Labour Force Survey - HLF > Labour Force Status by Sex by Age Group (Qrtly-Mar/Jun/Sep/Dec) Under Age Group, hold Ctrl and select: Aged 15-19 88 Aged 20-24 Total All Ages Under Sex, select Total sex. Under Observations, select Persons Employed in Labour Force. Select all the Time periods. At the bottom right of the screen, change Table on screen to csv file. Read this csv file into iNZight. You will need to remove unnecessary information from the csv first and rename the variables so that only the first row of the csv contains variable names. How to produce Time Series graph, Decomposition graph and Seasonal plot in iNZight? After you have imported your Time Series data click “Advanced” and then “Time Series…” The following command window will appear: Click if your dataset is Additive or Multiplicative. Then select the Time Series Plot option to produce the TOTAL graph, which includes the trend line. Then decompose the dataset using the Decompose option. This will produce the “Decomposition Total” graph. 89 Then select the Seasonplot option to produce the Seasonal plot and Multiplicative/Additive Seasonal Effects. How do you format your time variable so iNZight recognises it? INZight assumes that datasets are quarterly when producing a time series. If your time variable is not in the format the program recognises or if you have not included a quarter column, then you can supply the time structure-information through the module's command window. 1. Click Provide Time Information and then press Create. The following pop-up box will appear: 2. In Start Date enter the year in which the series started 3. In Frequency enter the number of seasons in a year (e.g. 52 for weekly data) 4. In Season Number enter the number of the season at which the series started (e.g. 3 for 3rd week of the year) Reinterpret the above if your data runs over weeks not years and the seasons are days of the week, or your data concerns days and hours in the day. 8. Number of flights by direction for Buenos Aires – Additive model 90 Total number of flights by direction for Buenos Aires: Time series plot for Total.number.of.flights.by.direction 40 30 20 10 0 2000 2005 2010 Time 91 2015 Decomposition of data:Total.number.of.flights.by.direction Trend Raw data 50 40 30 20 10 0 Seasonal Swing 10 5 0 -5 20 Residuals 15 10 5 0 -5 -10 2000 2005 Time 92 2010 2015 Seasonal plot for Total.number.of.flights.by.direction Additive seasonal effects 2002 2009 2001 2010 2011 2008 1999 2001 2006 2008 1999 2000 2006 2010 2007 2009 4 40 6 2000 2007 30 1998 2 2005 2005 1998 2012 2004 2011 1997 2003 0 20 2004 2003 0 -4 10 -2 2002 2015 2014 2013 Jan 2014 2013 2012 Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Month Feb Mar Apr May Jun Jul Month 93 Aug Sep Oct Nov Dec Number of Arrivals to New Zealand from Buenos Aires: The trend for the number of arrival flights to New Zealand from Buenos Aires for the, illustrated in the time Series plot and the Decomposition graph, reveal a decrease from 10 arrival flights 1997, to absolutely no flights since 2011. From 1997 to 2001 a significant increase can be seen. The trend then decrease, until 2003, where a gradual increase can be seen up until 2012/2013. The trend line then steeply decreases and remains constant from 2013 through until 2015. The decrease in frequency and cut in the number of arrival flights from Buenos Aires to New Zealand is most likely due to approvals and costs, such as jet fuel, number of passengers etc. According to The New Zealand Herald in December 2014, Air New Zealand announced it will be flying direct to Buenos Aires in December 2015.The airline had been weighing the option of providing a South American direct flight for about six months, therefore, when recent falls in jet fuel prices occurred, launching the route became a more attractive prospect. Looking at the seasonal pattern section of the Decomposition graph, the amplitude seems to be regular, suggesting that the number of arrival flights from Buenos Aires is somewhat effected by seasonality. The variation of the seasonal pattern ranges between -2 and 4. The seasonal component displayed in the Multiplicative seasonal effects graph shows the highest peak in January, followed by a slight decrease in February, however remaining constant until March. This is most likely due to the fact that these month mark New Zealand’s warmest climates, attracting many visitors as well as the Christmas-New Year, Easter and school holiday periods. From March to April, a relatively steep increase can be seen, further decreasing until reaching its minimum in June. The seasonal component then increases slightly in July, before decreasing slightly again in August. From August through till December, a steep increase can be seen. The seasonal plot shows a record high number of flight arrivals in January 2002. Further investigation should be carried out to identify the cause of this. The plot also highlights the massive decrease of flights in 2012, specifically in July, whereby, it is evident that from this point onwards, flight arrivals from Buenos Aires to New Zealand were cut. In addition, the residuals in the Decomposition graph reveal no particular pattern since 1997 and have varied between 10 and -5. It is small compared to the trend, however it is large compared to the seasonal component. Time series plot for Arrivals 20 15 10 5 0 2000 2005 2010 Time 94 2015 Decomposition of data:Arrivals Trend Raw data 25 20 15 10 5 0 Seasonal Swing 4 2 0 -2 10 Residuals 5 0 -5 2000 2005 Time 95 2010 2015 Seasonal plot for Arrivals Additive seasonal effects 3 2002 2008 2001 2011 2009 20 2000 2010 2009 2 2001 2000 1999 2006 2010 2007 1998 2007 15 1999 2006 2008 1 2005 1998 2005 2012 2004 1997 2003 2011 2003 0 10 2004 0 -2 5 -1 2002 2015 2014 2013 Jan 2014 2013 2012 Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Month Feb Mar Apr May Jun Jul Month 96 Aug Sep Oct Nov Dec Number of Departures from New Zealand to Buenos Aires: The trend for the number of departure flights to Buenos Aires from New Zealand, illustrated in the time series plot and the Decomposition graph, reveal the exact same trend as the number of arrival flights. Looking at the seasonal pattern section of the Decomposition graph, the amplitude seems to be regular, suggesting that the number of departure flights from New Zealand is somewhat effected by seasonality. The variation of the seasonal pattern ranges between -2 and 6. The seasonal component displayed in the Multiplicative seasonal effects graph shows the highest peak in January, followed by a slight decrease in February, however remaining constant until March. This is most likely due to the fact that during January and February marks Buenos Aires' summertime and peak tourist season. From March to April, a relatively steep increase can be seen, further decreasing until reaching its minimum in June. This is most likely due to the off-season for Buenos Aires starts in June and ends in August and is usually filled with rainy days, cold winter temps, and few tourists. The seasonal component then increases slightly in July, before decreasing slightly again in August. From August through till December, a steep increase can be seen. Like the seasonal plot for arrival flights, a record high is seen in January 2002. Further investigation should be carried out to identify the cause of this. The plot also highlights the massive decrease of flights in 2012, specifically in July, whereby, it is evident that from this point onwards, flight departures from New Zealand to Buenos Aires were cut. In addition, the residuals in the Decomposition graph reveal no particular pattern since 1997 and have varied between 10 and -5. It is small compared to the trend, however it is large compared to the seasonal component. Time series plot for Departures 25 20 15 10 5 0 2000 2005 2010 Time 97 2015 Decomposition of data:Departures Trend Raw data 25 20 15 10 5 0 6 Seasonal Swing 4 2 0 -2 10 Residuals 5 0 -5 2000 2005 Time 98 2010 2015 Additive seasonal effects 2000 3 25 Seasonal plot for Departures 2008 2011 2010 2009 20 1999 2002 2001 2009 2 2001 2006 2000 1999 2008 2007 2006 2010 2007 15 1998 2005 2011 1998 2004 2012 1997 2003 1 2005 10 0 2004 2002 2015 2014 2013 2014 2013 2012 0 -2 5 -1 2003 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Month Feb Mar Apr May Jun Jul Month 99 Aug Sep Oct Nov Dec COMPARISON OF THE SERIES: The actual values of the number of arrival flights and departure flights are relatively the same. This is expected as most aircrafts will do return trips to their country of residence. The trends for both series as well as the total are similar, all showing the significant event of the decrease in frequency and cut in the number of flights from mid-2012 onwards for Buenos Aires and New Zealand, most likely due to approvals and costs, such as jet fuel, number of passengers etc. The seasonal component for both series, as well as the total, displayed in the Multiplicative seasonal effects graphs show the highest peak in January, followed by a slight decrease in February, however remaining constant until March. This is most likely due to the fact that these month mark Buenos Aires' and New Zealand’s summertime and peak tourist season. All seasonal plots shows a record high number of flights by direction in January 2002. Further investigation should be carried out to identify the cause of this. The plot also highlights the massive decrease of flights in 2012, specifically in July, whereby, it is evident that from this point onwards, flights to and from Buenos Aires were cut. In addition the irregular for both variables as well as the total, are larger than the seasonal pattern, suggesting that for all the variables, the irregular has a larger influence on the data then the seasonal component. ACCESSING THE DATA How do I access the International Travel and Migration Statistics? This series is found in the International Travel and Migration section of the Statistics New Zealand website: www.stats.govt.nz (Statistics NZ Home > Browse for statistics > Population > Migration > International Travel and Migration) How do I get this particular data set and upload it to iNZight? On the Statistics New Zealand website you will find Quick links and Infoshare (www.stats.govt.nz/infoshare). Use the Browse tab. Select Tourism > International Travel and Migration-ITM > Number of flights by direction and overseas port (Monthly) Under Travel Direction, Select all variables. Under Overseas Port, select Buenos Aires. Select all the Time periods. At the bottom right of the screen, change Table on screen to csv file. 100 Read this csv file into iNZight. You will need to remove unnecessary information from the csv first and rename the variables so that only the first row of the csv contains variable names. Additional information about this dataset: Flights are only counted if at least one individual on the flight is recorded as an arrival or a departure. Although most flights are scheduled commercial services, the numbers may also include charter services, private aircraft, cargo aircraft and other special flights. Flights may visit more than one overseas port, and will be counted under each. Hence, the sum across overseas ports may differ from the total number of flights. Useful links: The New Zealand Herald, “Air NZ unveils new flights to Buenos Aires”- http://www.nzherald.co.nz/business/news/article.cfm?c_id=3&objectid=11372895 “Best Times to Visit Buenos Aires”- http://travel.usnews.com/Buenos_Aires_Argentina/When_To_Visit/ Jet Fuel Monthly Price - New Zealand Dollar per Gallon- http://www.indexmundi.com/commodities/?commodity=jet-fuel&months=360&currency=nzd How to produce Time Series graph, Decomposition graph and Seasonal plot in iNZight? After you have imported your Time Series data click “Advanced” and then “Time Series…” The following command window will appear: Click if your dataset is Additive or Multiplicative. Then select the Time Series Plot option to produce the TOTAL graph, which includes the trend line. Then decompose the dataset using the Decompose option. This will produce the “Decomposition Total” graph. 101 Then select the Seasonplot option to produce the Seasonal plot and Multiplicative/Additive Seasonal Effects. How do you format your time variable so iNZight recognises it? INZight assumes that datasets are quarterly when producing a time series. If your time variable is not in the format the program recognises or if you have not included a quarter column, then you can supply the time structure-information through the module's command window. 1. Click Provide Time Information and then press Create. The following pop-up box will appear: 2 .In Start Date enter the year in which the series started 3. In Frequency enter the number of seasons in a year (e.g. 52 for weekly data) 4 .In Season Number enter the number of the season at which the series started (e.g. 3 for 3rd week of the year) Reinterpret the above if your data runs over weeks not years and the seasons are days of the week, or your data concerns days and hours in the day. 102 Acknowledgements This resource has been supported by the Statistical Education and Statistical Methods team at Statistics New Zealand. In developing the examples presented here, helpful input and feedback from Emma Mawby, Senior Analyst in Statistical Education was provided. In addition, we would like to accredit Hayden Rebel and Andrew Richens from Statistical Methods for reviewing this resource. Te Aomihia Walker 103
