Social Research Methods: Qualitative and Quantitative Approaches, 5e • This multimedia product and its contents are protected under copyright law. The following are prohibited by law: any public performance or display, including transmission of any image over a network; preparation of any derivative work, including the extraction, in whole or in part, of any images; any rental, lease, or lending of the program. Chapter 8: Qualitative and Quantitative Sampling • Introduction • Nonprobability Sampling. • Probability Sampling. Introduction • Sampling = collecting data from only some of the cases about which one will generalize • In quantitative research, the concern is usually to draw a representative sample from a large population • In qualitative research, the focus may be on important cases or on “typical” cases, but fewer claims are made about representativeness There are two basic types of sampling • Probability: every case has some known, non-zero chance of being included in the sample – Random sampling: each element has equal chance of selection – But random is only one type of probability sampling • Nonprobability: no such calculation can be made Nonprobability Sampling • Only appropriate for research where drawing a representative sample from a population is not important • Typically, this means qualitative research, where cases are selected for some other reason(s) Types of Nonprobability Sampling (see p. 211, table 8.1) • Haphazard or Convenience – select cases that are easily studied, usually because they’re nearby – Text says not usually a good strategy (Why not?) – However, can be appropriate if nearby cases are likely to be useful (typical, exceptional, etc.) and if researcher can establish some basis for generalizing and contrasting • Quota – Identify relevant categories or types – Decide how many of each to study – Text says better than haphazard (Why?) Nonprobability Sampling continued… • Purposive or Judgmental Select cases for one of three specific reasons: – Unique cases that are very informative – Members of hard-to-reach categories – Identify particular types of cases for further study • Snowball – Ask each selected case to refer you to another – Useful in studying a network or group of connected cases Nonprobability Sampling continued… • Deviant Case – Select cases based on their unusualness or difficulty of finding • Sequential – Similar to purposive – select cases for a particular reason – Sample until ‘saturate’ – have enough information or variety of cases • Theoretical – Select cases according to theory – Text says evolves during grounded theory development – Can also be used in reconstructing theory Probability Sampling • • • • Formal Definition Population, Element, Sampling Frame Why Random Sampling? Types of Probability Samples – – – – – Simple random Systematic Stratified Cluster Random-Digit Dialing • Hidden Populations • Sample Size • Drawing Inferences Probability Sampling • Formal definition – every case has some known, non-zero chance of being included in the sample – Random sampling: each element has equal chance of selection – But random is only one type of probability sampling • Usually more difficult (time, money, etc.) than nonprobability – So, why do it? – More likely to generate a representative sample – Is the only type of sampling that allows researcher to estimate sampling error – This means you can use sample statistic to estimate population parameter, with some specified degree of confidence that true population parameter is within a specific range Probability Sampling • • • • Formal Definition Population, Element, Sampling Frame Why Random Sampling? Types of Probability Samples – – – – – Simple random Systematic Stratified Cluster Random-Digit Dialing • Hidden Populations • Sample Size • Drawing Inferences Populations, Elements, and Sampling Frames • Populations (or universes) - sets or pools of elements from which a sample is to be selected. • Elements - units of analysis such as persons, groups, organizations or agencies. Terminology, continued • Sampling Frame – list of population elements – intended to be exhaustive, contain no duplicates or foreign or missing elements. – Specifies unit being sample, geographical location, temporal boundaries • Start with population; precise definition creates a “target population” • Sampling ratio – what percentage of population elements will be selected for the sample (those actually studied) Terminology, continued • Population parameter: any characteristic of the population • Sample statistic: the relevant information from the sample, used to estimate the parameter Famous mistake – 1936: the Literary Digest predicts Alf Landon defeats Roosevelt for President • Main problem was a mistake in sampling – Despite large sample size – Sampling frame did not adequately represent the target population • Target population? – Voters • Sampling frame was based on automobile registrations and telephone directory – What is wrong with this sampling frame? Probability Sampling • • • • Formal Definition Population, Element, Sampling Frame Why Random Sampling? Types of Probability Samples – – – – – Simple random Systematic Stratified Cluster Random-Digit Dialing • Hidden Populations • Sample Size • Drawing Inferences Why Random Selection? • Random selection – A process that generates a mathematically random result – no pattern – Therefore, can assume that no human bias exists in the selection process. – Therefore, sample is more likely to be representative of the population Probability Sampling • • • • Formal Definition Population, Element, Sampling Frame Why Random Sampling? Types of Probability Samples – – – – – Simple random Systematic Stratified Cluster Random-Digit Dialing • Hidden Populations • Sample Size • Drawing Inferences Types of Probability Samples • • • • • Simple random Systematic Stratified Cluster Random-Digit Dialing Simple Random Sampling • Easiest probability sample to draw • Can think of in terms of drawing marbles from a jar, bingo numbers, etc. • In social science research, involves numbering the elements in a sampling frame and choosing numbers from a random number table Random sampling is the basis for understanding: • Sampling distribution: – Shows distribution of sample statistics for a number of independently-drawn samples • See box 8.2, p. 220 • Confidence intervals – A range around sample statistic, usually expressed as the statistic plus or minus some number – Within this range, researcher is confident to some specified level (usually 95% or higher) that population parameter is within this range Systematic Sampling • Start with a sampling frame • taking every xth case after a random start – i.e.: if 3 cases were needed out of 30, we could take each 10th case after selecting the first case randomly • Saves time relative to simple random sample • Can generate an unrepresentative sample, if – Elements in population are organized in some patterned way – E.g.: list of married heterosexual couples, with man’s name first; an even-numbered sampling interval will produce a sample of only one gender Stratified sampling • Useful when population contains different groups (strata) that are (likely to be): – Different from each other – Relatively homogenous internally • In this case, it may be best to – Separate the population into strata – Sample separately from each • Data analysis: – First conduct separately for each stratum – Then combine for overall sample statistics • Disproportionate sampling: in some cases, it is wise to use different sampling ratios for different strata; typically higher ratios for smaller strata Cluster Sampling • Useful when – it is hard to construct a sampling frame • Especially because of large size and geographic spread – E.g. college students in the United States – E.g. study of Iraqi civilian casualties – population elements are clustered • In first example: on college campuses • In second example: in cities • Can do multistage cluster sampling – First example: can do by state, then by college – Second example: city, then household – Each stage introduces more possible biases than does simple random • Can combine cluster and stratified Random-Digit Dialing • used with the general public when interviewing by telephone • Published directories miss those without phones, recently-moved, unlisted numbers, wireless numbers Probability Sampling • • • • Formal Definition Population, Element, Sampling Frame Why Random Sampling? Types of Probability Samples – – – – – Simple random Systematic Stratified Cluster Random-Digit Dialing • Hidden Populations • Sample Size • Drawing Inferences Hidden Populations • Are hard to reach people who may not want to be identified • may be able to be sampled by combining both qualitative and quantitative methods such as snowball sampling or asking currently identified persons to recruit others who are similar. How Large Should a Sample Be? • Researchers can use a formula based on: – Desired confidence level (usually 95%) – how much error she or he can tolerate (usually between 2–5%) – Size and estimated variability (heterogeneity) of the population • More common is to use “rule of thumb” – Based on: past experience, the number of variables being examined and the number of hypotheses being tested – See text p. 232 for some common sampling ratios Drawing Inferences • Inferential statistics: using sample statistics to make inferences about population parameters • The logic of sampling is similar to the logic of measurement: Population . . . . . . . Sample Concept . . . . . . . . . Measure See figure 8.5, page 234