Chapter 13 Sampling: quantitative and qualitative CONTENTS • • • • • Samples and populations Representativeness Sample size Weighting Sampling for qualitative research Samples and populations • Population: – Total category of subjects that is the focus of attention in a particular research project (can be non-human) • Sample: – A number of subjects drawn from the population • Two key issues: 1. What procedures must be followed to ensure that the sample is representative of the population? 2. How large should the sample be? A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Representativeness • Achieved by Random sampling: • A systematic selection process which ensures that all members of the population have an equal chance of inclusion in the sample • Designed to ensure representativeness • An unrepresentative sample is: biased • How is random sampling achieved in practice? A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Sampling for household surveys • Ideally – Eg. 10 million population – sample of 1000: all 10 m. names put in a drum and 1000 drawn. • In practice: – For national/regional surveys – multi-stage sampling used 1. Select states/regions 2. Within state/region select local government areas (lga) or constituencies/electorates 3. Within lgas or constituencies/electorates for face-to-face interviews select streets (telephone surveys select numbers at this point) 4. Select ‘clusters’ of 10-15 houses A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Sampling for telephone surveys • Telephone numbers selected at random from telephone directory • For large-scale surveys: automated by ComputerAided telephone Interviewing (CATI) • Requires access to electronic directory with residential/business numbers identified • No directories for mobile phones • For household and telephone surveys: select person in household randomly: eg person with next birthday A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Sampling for site/user/visitor surveys 1. ISUM: Interviewer stationary – user mobile: – eg. interviewing at entrance/exit – Sample by selecting: ‘next person to pass entrance/exit point 2. USIM: User stationary – interviewer mobile – eg. interviewing people on a beach – Interviewers should have a set route/rules to follow – eg. ‘interview every third person/group’ 3. Handouts – Handing out questionnaires to (all) visitors for self-completion – Not generally recommended unless closely supervised – generally very poor response rates A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Sampling for street/quota surveys • Can be used when data are available on key characteristics of population: – age/sex structure of a community – from Census • Interviewing target numbers determined by population characteristics – Eg. if Population Census indicates 12% retired: if overall sample size is 200: interview 24 retired people A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Sampling for mail surveys • Sample from mail-out list • 100% sample often used A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Sampling for complex events and destination surveys • Different components will conform to above guidelines – mostly site surveys • Problem lies in combining data from different sources for an overall result, if required. A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Sampling/ random assignment for experimental research • Samples of subjects often ‘convenience’ samples – eg. Students • Assignment to control and experimental groups: – Use of random numbers – Online: eg. www.random.org A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Sample size • Required sample size is not related to population size (except for small populations – see later) • Criteria: o o o The required level of precision in the results The level of detail in the proposed analysis The available budget A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Level of precision – confidence intervals • A statistic (finding) from a sample survey is an estimate of the population statistic • In a randomly drawn sample the sample value has a certain probability of being in a certain range either side of the population value – Eg. 95% probability of being within 2 ‘standard errors’ • See ‘Normal distribution’ – Theoretical: imagine drawing lots of samples: some would be accurate, some not – Discussed further in Ch. 17 A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Normal curve (Fig. 13.1) NUMBER OF SAMPLES 95% 2.5% -4 -3 Standard errors 2.5% -2 -1 +1 Popn Value +2 +3 +4 Standard errors A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Confidence intervals (CIs) (Table 13.1: Extract) Sample size (N) Percentages found from sample (‘results’) 50% 40/60% 30/70% 20/80% 10/90% 5/95% 2/98% 1/99% Confidence intervals (CIs) + % 500 +4.4 +4.3 +4.0 +3.5 +2.6 +1.9 +1.2 +0.9 A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Confidence intervals (CIs) (Table 13.1) Sample size (N) Percentages found from sample (‘results’) 50% 40/60% 30/70% 20/80% 10/90% 5/95% 2/98% 1/99% Confidence intervals (CIs) + % 500 +4.4 +4.3 +4.0 +3.5 +2.6 +1.9 +1.2 +0.9 So CI for 20% finding is 30% +4.0 = a range of: 26.0% to 34.0%. CI is not related to population size. NB. • CI for p = CI for 100-p – eg. CI is the same for 40% and 60% • CI for 50% is the largest in absolute terms • This table refers to 95% probability CIs – others can be calculated – eg. 99% A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Confidence intervals (CIs) contd (Table 13.1) Sample size (N) Percentages found from sample (‘results’) 50% 40/60% 30/70% 20/80% 10/90% 5/95% 2/98% 1/99% Confidence intervals (CIs) + % 500 +4.4 +4.3 +4.0 +3.5 +2.6 +1.9 +1.2 +0.9 2000 + 2.2 +2.1 +2.0 +1.7 +1.3 +1.0 +0.6 +0.4 So to halve the CI it is necessary to increase the sample fourfold. A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Confidence intervals (CIs) contd (Table 13.2) Table 13.1 can be changed to present necessary sample size for a given CI – see Table 13.2 Percentages found from sample (‘results’) Conf. Interval 50% 40/60% 30/70% 20/80% +1 9600 9216 8064 6144 +2 2400 2304 2016 +4 600 576 +8 150 144 10/90% 5/95% 1/99% 3456 1824 380 1536 864 456 * 504 384 216 114 * 126 96 53 * * Necessary sample sizes A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Suggested appendix on sample size and CIs • See Appendix 13.1 • table indicating levels of Cis • statement indicating that they have been taken into account A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Detail of proposed analysis Sample size % CI Range, % Comment Survey with sample of 200 200 Bowling 20 +5.5 14.5 – 25.5 Tennis 30 +6.3 23.7 – 36.3 Ranges overlap Survey with sample of 500 500 Bowling 20 +3.5 16.5 – 23.5 Tennis 30 +4.0 26.0 – 34.0 Ranges do not overlap A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Budget • Key issue: halving the CI requires 4-fold increase in sample size – Eg. N = 250 CI for 50% = +6.2 Survey Cost = 200 x $20 = $5000 N = 1000 CI for 50% = +3.1 Survey Cost = 1000 x $20 = $20,000 • If resources not available for adequate sample size, consider: – Pilot/exploratory study – Qualitative study A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Confidence intervals and population estimates (Table 13.3) Item Source Number Population Census 500,000 Sample Survey 1,000 % visiting facility/year Survey 12% % confidence interval Table 13.1 + 2.0% Estimated no. persons 12% of 500,000 60,000 CI in no. persons + 2% of 500,000 + 10,000 CI as % of persons (10,000/60,000) x 100 + 16.7% Frequency of visit/year Survey Estimated total visits Calc.: (12% of 500,000) x 2.5 150,000 CI in no. visits Calc: (2% of 500,000) x 2.5 +25,000 CI in % visits Cals.: (25,000/150,000) x 100 2.5 +16.7% A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Sampling for small populations • CIs are affected by population size if population is below about 50,000 • See Table 13.3 A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Sample size & population size: small pop’ns (Table 13.3) Population size Minimum sample size to achieve CI of +5% or +1% on a sample finding of 50% +5% +1% Infinite 384 9602 5 million 384 9584 1 million 384 9511 500,000 384 9422 100,000 383 8761 50,000 381 8056 10,000 370 4899 5000 357 3288 1000 278 906 100 80 99 A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Weighting (Tables 13.4 & 13.5) Example: one survey at a site Time 9-11 am 11.01-1 pm 1.01-3 pm 3.01-5 pm Total # of Interviews 10 12 11 12 45 % 22.2 26.7 24.4 26.7 100.0 Actual # of users (counts) 25 240 110 60 435 % 5.7 55.2 25.3 2.7 100.0 Sample does not reflect the pattern of use A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Weighting contd (Table 13.5) Time Source: 9-11 am 11.01-1 pm 1.01-3 pm 3.01- 5pm Total A No. of Interviews Survey 10 12 11 12 45 B No. of Users Counts 25 240 110 60 435 C Weighting Factors B/A 2.5 20.0 10.0 5.0 D Weighted Sample No. CxA 25 240 110 60 435 A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Sampling for qualitative research • Number of subjects generally be small, but: – sampling process is still important – should be fully described in research report • A range of approaches is possible A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge Sampling for qualitative research contd (Table 13.6) Method Convenience Criterion Homogeneous Opportunistic Maximum variation Purposeful Snowball Stratified purposeful Characteristics Conveniently located persons or organisations Selected on key criterion - eg. age-group. Deliberately homogeneous group: eg. universityeducated male cyclists aged 20-30. Taking advantages of opportunities as they arise - eg. a major sporting event taking place locally. Deliberately studying contrasting cases. Opposite of 'homogeneous'. Similar to 'criterion' but may involve other considerations, such as 'maximum variation', typicality. Interviewees source of suggestions for contacts. A range of cases based on set criteria, eg. representatives of a range of age-groups or nationalities. A. J. Veal & S. Darcy (2014) Research Methods for Sport Studies and Sport Management: A practical guide. London: Routledge