3. When a researcher wants to identify particular types of cases for in-depth investigation; purpose less to generalize to larger population than to g

Transcription

1 Chapter 7: Qualitative and Quantitative Sampling Introduction Quantitative researchers more concerned with sampling; primary goal to get a representative sample (smaller set of cases a researcher selects from large pool and generalizes to population) and tend to use sampling based on theories of probability (called probability sampling) Using probability/random sampling has two motivations: 1. Save time and cost and 2. Accuracy Census: an attempt to count everyone in a target population (takes place in Canada every 5 years) Qualitative researchers focus on how the sample or small collection of cases illuminates key features of social life; purpose of sampling is to collect cases, events or actions that clarify and deepen understanding o Focus on finding cases that will enhance what the researchers learn about processes of social life in specific context and use nonprobability sampling Nonprobability Sampling Non-random sample: type of sample in which the sampling elements are selected using something other than a mathematically random process Rarely determine sample size in advance and have limited knowledge about large group/population from which sample is taken Select cases gradually with specific context of case determining whether it is chosen Types of Nonprobability samples: Haphazard Get any cases in any manner that is convenient Quota Get a pre-set number of cases in each of several predetermined categories that will reflect diversity of population, using haphazard methods Purposive Get all possible cases that fit particular criteria, using various methods Snowball Get cases using referrals from one or few cases, and then referrals from those cases, and so on Sequential Get cases until there is no additional information/new characteristics (*often used with other sampling methods) Haphazard, Accidental, or Convenience Sampling Haphazard sampling: a type of non-random sample in which the researcher selects anyone he happens to come across Can produce ineffective, unrepresentative samples and not recommended Cheap and quick but many systematic errors I.e. person on the street interviews seen on TV Quota Sampling Defn: type of non-random sample in which the researcher first identifies general categories into which cases or people will be selected, then he selects predetermined number of cases in each category Researcher can ensure that some differences are in the sample (i.e. age) Researchers use haphazard sampling once the quota samples fixes the categories and number of cases in each category Purposive Sampling Defn: researcher uses wide range of methods to locate all possible cases of a highly specific and difficult-to-reach population Used in situations in which expert uses judgment in selecting cases with specific purpose in mind Researcher never knows whether the cases selected represent the population Appropriate in three situations: 1. Researcher uses it to select unique cases that are especially informative 2. To select members of difficult-to-reach, specialized population; i.e. researcher wants to study prostitutes so he finds different ways to find as many to include in his study as possible (places where they solicit, social groups they interact with or police who work with prostitutes)

2 3. When a researcher wants to identify particular types of cases for in-depth investigation; purpose less to generalize to larger population than to gain deeper understanding of types Deviant case sampling: type of non-random sample, especially used by qualitative researchers, in which a researcher selects unusual or nonconforming cases purposely as a way to provide greater insight into social processes or a setting o Seek cases that differ from dominant pattern or that differ from predominant characteristics of other cases o Goal is to locate collection of unusual, different, or peculiar cases that are not representative of the whole o I.e.. Researcher studying high school dropouts Snowball Sampling Defn: type of non-random sample in which the researcher begins with one case, then, based on information about interrelationships form that case, identifies other cases, and then repeats the process again and again Also called network, chain referral or reputational sampling Method of identifying and sampling the cases in a network Social researchers often interested in interconnected network of people or organizations Crucial feature is that each person or unit connected with another through direct/indirect linkage Sociogram: diagram or map that shows the network of social relationships, influence patterns or communication paths among group of people or units Also use snowball sampling in combination with purposive sampling as in case of Albanese (2006) in qualitative study of women in Quebec whose children were in provincial childcare Sequential Sampling Defn: type of non-random sample in which a researcher tries to find as many relevant cases as possible, until time, financial resources, or his energy are exhausted, and there is no new information or diversity from the cases Information is gathered until marginal utility, or incremental benefit for additional cases, levels off or drops significantly Theoretical sampling: an iterative sampling technique associated with the grounded theory approach in which the sample size is determined when the data reach theoretical saturation; continue to collect data until no new information emerges Theoretical saturation: a term associated with grounded theory approach that refers to the point at which no new themes emerge from the data and sampling is considered complete Probability Sampling Populations, Elements, and Sampling Frames Researcher draws sample from larger pool of cases, or elements Sampling element: name for a case or single unit to be selected; unit of analysis in population o Can be a person, group or organization Large pool is the population (name for large general group of many cases from which researcher draws sample and which is usually stated in theoretical terms); can also be called universe Target population: name for large general group of many cases from which a sample is drawn and which is specified in very concrete terms; specific pool of cases that he wants to study Sampling ratio: number of cases in the sample divided by the number of cases in the population or the sampling frame, or the proportion of the population in the sample; ratio of the size of the sample to the size of the target population Population is an abstract concept, cant be frozen at any time to measure it accurately o Therefore, the researcher needs to estimate the population; researcher operationalizes a population by developing specific list that closely approximates all the elements in population o Sampling frame: list of cases in a population, or the best approximation of it (i.e. telephone directories, tax records)

3 o Good sampling frame crucial to good sampling Population parameter: characteristic of the entire population that is estimated from a sample; determined when all elements in population are measured o Never known with absolute accuracy for large populations o Statistic: numerical estimate of population parameter computer from a sample Why Random? Probability relies on random processes Random: refers to process that generates mathematically random result; selection process operates in truly random method and researcher can calculate probability of outcomes o Each element has equal probability of being selected Sampling error: how much a sample deviates from being representative of the population; deviation between sample results and a population parameter due to random processes Margin of error: estimate about the amount of sampling error that exists in a survey s results; estimate about the amount of sampling error that exists in survey s results Random sample: type of sample in which the researcher uses a random number table or similar mathematical random process so that each sampling element in the population will have an equal probability of being selected Types of Probability Samples Simple Random Sampling Defn: type of random sample in which a researcher creates a sampling frame and uses a pure random process to select cases; each sampling element in the population will have an equal probability of being selected Researcher develops accurate sampling frame, selects elements from sampling frame according to mathematically random procedure, then locates exact element that was selected for inclusion in the sample After numbering all elements in sampling frame, a researcher uses list of random numbers to decide which elements to select Needs as many random numbers as there are elements to be sampled Can get random numbers from random-number table (list of numbers that has no pattern in it and that is used to create random process for selecting cases and other randomization purposes) o Random number tables available in most statistics and research methods books Sampling distribution: distribution created by drawing many random samples from the same population; distribution of different samples that shows frequency of different sample outcomes from many separate random samples Central limit theorem: law-like mathematical relationship stating that whenever many random samples are draw from a population and plotted, a normal distribution is formed, and the centre of such a distribution for a variable is equal to its population parameter; tells us that number of different random samples in sampling distribution increase toward infinity, the pattern of samples and the population parameter become more predictable o With large number of random samples, the sampling distribution forms normal curve and midpoint of curve approaches population parameter as number of samples increase o Allows researcher to generalize from sample to population without actually having to take many different samples Most random samples will be close to population and one can calculate probability of particular sample s being inaccurate o Researcher measures chance that particular sample is off or unrepresentative by using information from sample to estimate sampling distribution Confidence interval: range of values, usually a little higher and lower than specific value found in a sample, within which a researcher has a specified and high degree of confidence that the population parameter lies o Range around specific point used to estimate a population parameter

4 o Used because statistics of random processes do not let researcher predict exact point but they let researcher say with high level of confidence (i.e. 95%) that the true population parameter lies within a certain range Systematic Sampling Defn: type of random sample in which a research selects every nth (i.e. 9 th ) case in the sample frame using a sampling interval First step is to number each element in sampling frame Researcher calculates sampling interval and the interval becomes his/her quasi-random selection method Sampling interval: inverse of the sampling ratio, which is used in systematic sampling to select cases. The sampling interval (i.e. 1 in n, where n is some number) tells the researcher how to sample elements from sampling frame by skipping elements in the frame before selecting one for the sample; tells the researcher how to select elements from sampling frame by skipping elements in frame before selecting one for the sample Simple random sample and systematic sample yield virtually same results in most cases other than one o Systematic sampling cannot be substituted for simple random sampling when the elements in a sample are organized in some kind of cycle or pattern Stratified Sampling Defn: type of random sample in which the researcher first identifies a set of mutually exclusive or exhaustive categories, then uses a random selection method to select cases for each category Researcher first divides population into subpopulations (strata) on basis of supplementary information o Then researcher draws random sample or systematic sampling In general stratified sampling produces samples that are more representative of population that simple random sampling if the stratum information is accurate Used when stratum of interest is small percentage of a population and random processes could miss the stratum by chance Cluster Sampling Defn: type of random sample that uses multiple stages and is often used to cover wide geographic areas in which aggregated units are randomly selected; samples are then drawn from the sampled aggregated units, or clusters Addresses two problems: researchers lack good sampling frame from dispersed population and cost to reach sampled element is very high Instead of using single sampling frame, researchers use sampling design that involves multiple stage and clusters Cluster: unit that contains final sampling elements but can be treated temporarily as a sampling element itself Researcher first samples clusters each of which contains elements and then draws second sample from within clusters selected in the first stage of sampling Researcher draws several samples in stages in cluster sampling Stage 1: random sampling of big clusters Stage 2: random sampling of small clusters within each selected big cluster Stage 3: sampling of elements from within that sampled small clusters Usually less expensive than simple random sampling but less accurate; each stage introduces sampling errors Researcher has to decide on number of clusters and number of elements within each cluster Probability Proportionate to Size (PPS): two methods of cluster sampling o One method is proportionate ^ because size of each cluster is the same o More common, however, for cluster sizes to be different o Defn: an adjustment made in cluster sampling when each cluster does not have the same number of sampling elements Random-Digit Dialing (RDD)

5 Defn: method of randomly selecting cases for telephone interviews that uses all possible telephone numbers as a sampling frame General public interviewed by phone Three kinds of people misses: those without landline phones, people who recently moved and people with unlisted numbers Sampling element is the phone number, not the person or household Hidden Populations Defn: people who engage in secret, deviant, or concealed activities and who are difficult to locate and study i.e. illegal drugs users, sex workers and homeless people How large should a sample be? Depends on kind of data analysis required for the research, how accurate the sample has to be for the researcher s purposes, and on population characteristics The smaller the population, the larger the sampling ratio has to be for an accurate sample researcher s decision about best sample size depends on three things: degree of accuracy required, degree of diversity in population and number of different variables examined simultaneously in data analysis Drawing Inferences inferential statistics: branch of applied mathematics or statistics based on a random sample; lets researchers make precise statements about the level of confidence he has in the results of a sample being equal to the population parameter sample represents the population Conclusion probability sampling preferred by quantitative researchers because it produces sample that represents population and enables the researcher to use powerful statistical techniques