Fundamentals of Sampling Method Week 4 Research Methods & Data Analysis Dr. Mario Mazzocchi Research Methods & Data Analysis 1 Tutorials • Thursday 30th October 9-11 AG GL 20 (M. Mazzocchi) • Tuesday 4th November 11-1pm (H.Neeliah) • You may attend: – One (the most convenient for you) – Both (it may be very useful) – None (not really advised…) Dr. Mario Mazzocchi Research Methods & Data Analysis 2 Lecture outline • • • • • Key notions of statistics Simple random sampling Sampling error Sampling size Other sampling methods Dr. Mario Mazzocchi Research Methods & Data Analysis 3 Distributions • A set of values of a set of data together with their Count – Absolute frequencies – Relative frequencies (probabilities) 60 40 20 0 200.00 300.00 400.00 500.00 600.00 700.00 Amount spent Dr. Mario Mazzocchi Research Methods & Data Analysis 4 Relative and cumulate frequencies fi=ni/N Fi f1 f 2 i fi f h h 1 100% 8% 75% Perce nt Percent 6% 4% 50% 25% 2% 0% 200.00 300.00 400.00 500.00 600.00 700.00 200.00 300.00 Amount spe nt Dr. Mario Mazzocchi 400.00 500.00 600.00 700.00 Amount s pent Research Methods & Data Analysis 5 Distributions of random variables • The distribution of possible values together with their probabilities (probability density function, p.d.f.) Dr. Mario Mazzocchi Research Methods & Data Analysis 6 The normal (Gaussian) distribution • …is the distribution representing perfect randomness around a mean value • In statistics, the normal distribution play a key role in the theory of errors • The central limit theorem implies that “averaging” almost always give origin to a normal distribution (error on the average is random), provided that the number of observation is large (>40) Dr. Mario Mazzocchi Research Methods & Data Analysis 7 The normal distribution p 95% of values 0,025 0,025 m-1.96s Dr. Mario Mazzocchi m m+1.96s Research Methods & Data Analysis 8 The student-t distribution • When the parameter in the population has a normal distribution (with unknown variance), within the sample the parameter assumes a t distribution • The t-distribution is similar to the normal distribution, apart from having higher tailprobabilities • The bigger is the sample, the more similar the tdistribution is to the normal distribution • For samples with more than 30-40 units, the difference between the two distributions is negligible Dr. Mario Mazzocchi Research Methods & Data Analysis 9 The t-distribution x-ta/2sx Dr. Mario Mazzocchi x x+ta/2sx Research Methods & Data Analysis 10 ta/2 and za/2 – tabled values Level of confidence 99% 95% 90% Dr. Mario Mazzocchi t according to sample size a a /2 10 20 30 0.01 0.005 3.17 2.85 2.75 0.05 0.025 2.23 2.09 2.04 0.10 0.050 1.81 1.72 1.70 Research Methods & Data Analysis z 40 2.70 2.02 1.68 2.58 1.96 1.64 11 Population parameters (in a population of N elements) 1 N • Mean m xi N i 1 N 1 • Variance 2 s ( xi m )2 N i 1 • Standard deviation N 1 2 s s2 ( x m ) i N i 1 Dr. Mario Mazzocchi Research Methods & Data Analysis 12 Sampling • A sample is a subgroup of the population selected for the study • Sample statistics allow to make inference about the population parameters, through estimation and hypothesis testing • The sample space is a complete set of all possible results of the sampling procedure Dr. Mario Mazzocchi Research Methods & Data Analysis 13 Simple random sampling • Each element of the population has a known and equal probability of selection • Every element is selected independently from other elements • The probability of selecting a given sample of n elements is computable (known) • The Central Limit Theorem guarantees that for simple random samples with sample size (n) sufficiently large (>40), the sample mean in a S.R.S. follows the normal distribution Dr. Mario Mazzocchi Research Methods & Data Analysis 14 Sample statistics • Sample mean 1 n x xi n i 1 • Sample variance n 1 2 2 s ( xi x ) n 1 i 1 unbiasedness • Sample standard deviation n 1 2 2 s s ( xi x ) n 1 i 1 Dr. Mario Mazzocchi Research Methods & Data Analysis 15 Standard deviation and standard error • The standard deviation measures the variability of a given variable (e.g. X) within the population or sample • The standard error refers to the accuracy (variability) of the sample statistics (e.g. mean), i.e. the error due to the fact that the statistic is computed on a sample rather than on the population (sampling error) Dr. Mario Mazzocchi Research Methods & Data Analysis 16 Basic SRS sample statistics (unknown pop. variance) n Mean case x x i i 1 Proportion case (p) n n s 2 ( x x ) i i 1 n 1 sx s2 n Sample standard deviation of X Standard error of the mean/proportion s n p(1 p) n 1 sp p(1 p) n 1 ACCURACY of sample estimates Dr. Mario Mazzocchi Research Methods & Data Analysis 17 Finite population correction factor • For finite population (…i.e. all in social research), large samples (more than 10% of N) tend to overestimate the standard error of the sample mean (proportion) • In order to account for that, the following correction is necessary sx n s2 1 N n Dr. Mario Mazzocchi sp p(1 p) n 1 n 1 N Research Methods & Data Analysis 18 Level of confidence a and z parameter The level of confidence a refers to the probability that the true population mean falls in the identified confidence interval For the normal distribution, given a value of a, the corresponding za/2 values is tabulated a/2 x za / 2 sx a/2 x x za / 2 sx a=0.05 za/2 =1.96 Confidence interval for x at a level of confidence a Dr. Mario Mazzocchi Research Methods & Data Analysis 19 The t-distribution x-ta/2sx Dr. Mario Mazzocchi x x+ta/2sx Research Methods & Data Analysis 20 Confidence intervals • Calculate the sample mean • Decide a level of confidence (usually 95% or 99%) • Choose whether using the Student-t distribution or the Normal distribution • Compute the sample standard error • Define the lower and upper bound of the confidence interval Dr. Mario Mazzocchi Research Methods & Data Analysis 21 Exercise • Suppose that you have interviewed 20 students out of 200 in the agricultural building, asking them how much they paid for lunch yesterday • You get an average of £ 3.67 • The standard deviation is 1.25 • Compute the 95% confidence interval • Compute the 99% confidence interval Dr. Mario Mazzocchi Research Methods & Data Analysis 22 Determining sample size Factors influencing sample size (n): • Size of the population (N) • Variability of the population (s) • Desired level of accuracy (q) • Level of confidence (a) • Budget constraint Dr. Mario Mazzocchi Research Methods & Data Analysis 23 Simple random sampling: determining sample size • Relative sampling error (r.s.e) r ta / 2 sx nX n 1 N • Determining sampling size for a given r.s.e. (approximate formula) ta / 2 sx n0 rX Dr. Mario Mazzocchi 2 Research Methods & Data Analysis 24 The sampling design process 1. Define the target population, its elements and the sampling units 2. Determine the sampling frame (list) 3. Select a sampling technique • Sampling with/without replacement • Probability/Nonprobability sampling 4. Determine the sample size • Precision versus costs • The marginal value in terms of precision of additional sampling units is decreasing 5. Execute the sampling process Dr. Mario Mazzocchi Research Methods & Data Analysis 25 The sampling techniques • Probabilistic samples – Simple random sampling – Systematic sampling – Stratified sampling – Cluster sampling – Other sampling techniques • Nonprobabilistic samples – Convenience sampling – Judgmental sampling – Quota sampling – Snowball sampling Dr. Mario Mazzocchi Research Methods & Data Analysis 26 Representativeness • A sample can be considered as “representative” when it is expected to exhibit the average properties of the population Dr. Mario Mazzocchi Research Methods & Data Analysis 27 Selection bias • Improper selection of sample units (ignoring a relevant “control variable” that generate bias), so that the values observed in the sample are biased and the sample is not representative. Example: A survey is conducted for measuring goat milk consumption, but the interviewers just select people in urban areas, that on average drink less goat milk. Dr. Mario Mazzocchi Research Methods & Data Analysis 28 Simple random sampling • Each element of the population has a known and equal probability of selection • Every element is selected independently from other elements • The probability of selecting a given sample of n elements is computable (known) –Statistical inference is possible –It is easily understood Dr. Mario Mazzocchi –Representative samples are large and expensive –Standard errors are larger than in other probabilistic sampling techniques –Sometimes it is difficult to execute a really random sampling Research Methods & Data Analysis 29 Systematic sampling • A list of N elements in the population is compiled, ordered according to a specified variable – Unrelated to the target variable (similar to SRS) – Related to the target variable (increased representativeness) • A sampling size n is chosen • A systematic step of k=N/n is set • A random number s between 1 and N is extracted and represents the first element to be included • Then the other elements selected are s+k, s+2k, s+3k… –Cheaper and easier than SRS –More representative if order is related to the interest variable (monotone) –Sampling frame not always necessary Dr. Mario Mazzocchi –Less representative (biased) if the order is cyclical Research Methods & Data Analysis 30 Stratified sampling • Population is partitioned in strata through control variables (stratification variables), closely related with the target variable, so that there is homogeneity within each stratum and heterogeneity between strata • A simple random sampling frame is applied in each strata of the population – Proportionate sampling: size of the sample from each stratum is proportional to the relative size of the stratum in the total population – Disproportionate sampling: size is also proportional to the standard deviation of the target variable in each stratum –Gains in precision –Stratification variables may not be easily identifiable –Include all relevant subpopolation even if small –Stratification can be expensive Dr. Mario Mazzocchi Research Methods & Data Analysis 31 Cluster sampling • • 1. The population is partitioned into clusters Elements within the cluster should be as heterogeneous as possible with respect to the variable of interests (e.g. area sampling) A random sample of clusters is extracted through SRS (with probability proportional to the cluster size) – – 2a. All the elements of the cluster are selected (onestage) 2b. A probabilistic sample is extracted from the cluster (two-stage cluster sampling) –Reduced costs –Higher feasibility Dr. Mario Mazzocchi –Less precision –Inference can be difficult Research Methods & Data Analysis 32 Non probabilistic samples Dr. Mario Mazzocchi Research Methods & Data Analysis 33 Convenience sampling • Only “convenient” elements enter the sample –Cheapest method –Quickest method Dr. Mario Mazzocchi –Selection bias –Non representativeness –Inference is not possible Research Methods & Data Analysis 34 Judgmental sampling • Selection based on the judgment of the researcher –Low cost –Quick Dr. Mario Mazzocchi –Non representativeness –Inference is not possible –Subjective Research Methods & Data Analysis 35 Quota sampling 1. Define control categories (quotas) for the population elements, such as sex, age… 2. Apply a “restricted judgmental sampling”, so that quotas in the sample are the same of those in the population –Cheapest method –Quickest method Dr. Mario Mazzocchi –There is no guarantee that the sample is representative (relevance of control characteristic chosen) –Many sources of selection bias –No assessment of sampling error Research Methods & Data Analysis 36 Snowball sampling • A first small sample is selected randomly • Respondents are asked to identify others who belong to the population of interests • The referrals will have demographic and psychographic characteristics similar to the referrers –Lower costs –Low variability –Useful for “rare” populations Dr. Mario Mazzocchi –Inference is not possible Research Methods & Data Analysis 37