1. Sampling

2. Concerns 1)Representativeness of the Sample: Does the sample accurately portray the population from which it is drawn 2)Time and Change: Was the sample drawn during a period that represents the period in which conclusions are made [e.g. if public opinion is rapidly changing the sample might become non-representative very fast]. 3)Uncertainty of responses: and how to deal with them so as not to introduce bias into the representative nature of the sample.

3. TWO TYPES OF Sampling Methods 1. Non-Probability Sampling 2. Probability Sampling

4. Non Probability Samples Non Probability Samples We do not know the size of the population from which the sample was drawn, e.g. there is no list of ‘homeless people’ available. Therefore, we do not know how representative of the entire population the responses are, since we cannot control for social- demographic characteristics of the population. Generalizability to form ‘laws’ is therefore problematic but they offer greater insight compared to doing nothing.

5. Non Probability Non Probability 1. Convenience : easily available subjects- the least reliable of the non-probability sampling methods. 2. Purposive: as your understanding of the situation improves you concentrate on the group that is most relevant to your study, purposefully studying only those elements that belong to the group. – Purposefully looking for deviant cases is also an example, the exception reveals much about the generality and the conditions for its existence 3. Snowball: accidental, locating a few individuals and asking them to locate other members if their specific population. – Generalizability problems aside, this method of sampling reveals important insights into the group’s ‘social networks.’

6. Non Probability Non Probability 4. Quota: Draw up a matrix that lists the major characteristics of the target population. Then collect data purposefully from members that are identical to characteristics in various cells of the matrix. – Problem: How do you assess the accuracy of the quote frame- do those cells actually represent the characteristics of the target population – Even if your quota frame is accurate, the selected elements might not be representative of the general element belonging to that cell because they are non- randomly picked.

7. Non Probability Non Probability 5. Selecting Informants – An informant is a member of a group who can talk directly about the group, as against a respondent who provides information about himself/herself which you then use to learn about the group made up of respondents. – You want to select informants that are typical of elements of the group- in other words you must have a demographic profile of the group/target population – A marginal informant might have a non-typical view of the group and might limit your access to the group, you should therefore seek a typical member, or members that will be in a position to offer you access to typical members.

8. Non Probability Non Probability The problems with non-probability samples are: – Non generalizability of the results (advertent sampling bias) – Non representativeness of the target population (advertent sampling bias) – Greater researcher bias since the researcher determines who the elements of study are going to be (inadvertent sampling bias) – Benefits of non-probability samples Availability of respondents versus no respondents Great insight into processes, better than nothing and greater view of variability as against generalizability Good foundation for concept formation and exploratory studies

9. Probability Sampling Probability Sampling We do know the size of the population from which the sample was drawn. We do therefore have an idea about the representativeness of the sample, and its social-demographic characteristics. – In order to provide a useful description of the total population, the sample should contain the same variations that exist in the population

10. Probability Probability Sampling Bias: – those selected are not typical or representative of the larger population they have been chosen from. Representativeness of the sample: – If the aggregate characteristics of the sample closely approximate the same aggregate characteristics in the population – A sample is representative of the population from which it is selected if all members of the population have an equal chance of being selected in the sample – EPSEM (equal probability of selection method) is the way to ensure representativeness of the sample and implies random selection of the sample independent of any event or bias. Probability Samples: – Allow us to calculate sampling error and thus establish the degree to which the sample is representative – Are typically much more representative than non-probability samples

11. Probability Sampling Probability Sampling Terms to remember: Population, Study Population, Sampling Frame, Sampling Unit (element), Sampling Error. Population : The group about which we are making generalizations. E.g. If we are interested in studying diabetes patients in the U.S., our population (the one we will make generalizations about) will be: ALL diabetes patients in the United States. The population will not be ALL diabetes patients and it will not be All U.S. residents. Note: if generalizations are made about a group, the unit of analysis is the group, if generalizations are made about a nation state, the unit of analysis is the nation state. If generalizations are made about the world, the unit of analysis is the world.

12. Probability Sampling Probability Sampling Study Population: – The aggregation (or list) of elements from which the sample is actually drawn. E.g. After we have determined that the population about which we will make generalizations will be “ All diabetes patients that are residents of the U.S.” we will need a list of all such people, a list that has them all as an aggregate. That list will be the study population and the listing of elements is your Sampling Frame Element (or Sampling Unit) – Is the unit about which information is collected and provides the basis of analysis; E.g. the one man or woman chosen to be part of the sample from whom you collect information- also called the unit of observation. – A parameter is a summary description of some variable in the in a population (E.g. mean income or the age distribution of a population etc.).

13. Probability Sampling Probability Sampling Statistic: – The summary description of a variable in the sample selected. – The basic purpose of random sampling is that the sample statistic gives accurate estimates of the parameter of the total population from which the sample was drawn. – If we draw many ‘random samples’ from a population, the sample statistics of those samples will be distributed around the population parameter in a known way and probability theory tells us how closely the sample statistics are clustered around the true population value, i.e. we can estimate the Sampling Error, the degree of error expected for a given sampling design.

14. Probability Sampling Probability Sampling Confidence Levels: – In reality we do not know the population parameter – We also do not draw a large number of samples on one sample – Probability theory however tells us that 95% of the samples drawn would fall within two standard errors of the parameter and that 68% would fall within one standard error and that 99.9 % fall within three standard errors of the parameter. – So if you get a mean sample income of $10,000 per family, and you calculate the standard error ( standard deviation) to be $1000 you can say that you are 95% confident that the population parameter is between $8000 to $12000, i.e. + or – two standard errors. The $8000 to $12000 is the range of values between which the population parameter is supposed to be and is the Confidence Interval

15. Types of Sampling Design Types of Sampling Design 1. Simple Random Sample (SRS) – The basic sampling design assumed by all social research that relies on statistical computation of results. E.g. Develop a sampling frame listing all study population elements, assign a single number to all elements, you a computer generated list of random numbers to pick the sample you desire based on the sample size you have chosen. 2. Systematic Sampling – This is identical to simple random sampling (SRS), except you systematically pick every kth element (sampling interval= population size/sample size) instead of using a list of random numbers generated by the computer, after selecting the first element at random – A danger in this type of sampling is that if the list of elements is arranged in any way, it might produce a systematic bias in selection, e.g. if officers are listed first and then privates and you have a big sampling interval you might miss most of the officers and your sample might over represent privates etc.) – If implicit stratification (based on some variable)with a random start this can reduce sampling error by increasing homogeneity between elements.

16. Types of Sampling Design Types of Sampling Design 3. Stratified Sampling – Aims to reduce the sampling errors of simple random and systematic random samples There are two ways to reduce sampling errors, i) increase sample size, ii) make the population more homogeneous Stratified sampling tries to reduce sampling error by making the population more homogeneous through: i) identifying variables that produce heterogeneity in the population, ii) making homogeneous sampling frames based on those variables iii) picking random or systematic samples from each sampling frame based on their population percentage.

17. Types of Sampling Design Types of Sampling Design 4. Multistage Cluster Sampling – Used when populations cannot be easily listed for sampling – Involves primary sampling of larger units then a secondary sampling of their elements e.g. churches and then church members – Produces less accurate samples because every stage adds a sampling error. The way to control this is to increase homogeneity of the clusters, increase cluster (primary sampling) number and reduce elements in each cluster (secondary sampling) E.g. Stage 1- Identify city blocks and take a random sample of those block (primary sampling unit) Stage 2- For each block, list households in order (secondary sampling unit) Stage 3- From the household list select households to interview randomly.

18. Types of Sampling Design Types of Sampling Design 5. Probability Proportionate to Size (PPS) Sampling. – When clusters sampled are of different sizes, e.g. the city blocks or households in the previous example, we use PPS, each cluster is given a “chance of selection” equal to its size If for example we want 100 households out of 1000, the chance of selection of a household is 100/1000= 0.1, however different city blocks have different number of households, the larger ones have a greater chance of selection, so to fix this we reduce the elements selected from the larger to give the smaller an “equal probability of selection.” 6. Disproportionate Sampling and weighting This is used only when you want to concentrate on a particular sub population, so you over-sample from that group to have a sufficient number of cases to conduct your analysis.