Sampling Methods and types, Exams of Statistics

Download Sampling Methods and types and more Exams Statistics in PDF only on Docsity!

SAMPLING

Week 6 Slides ScWk 240

Purpose of Sampling

Why sampling? -­‐ to study the whole popula5on? A major reason studying samples rather than the whole group is that the whole group is so large that studying it is not feasible. Example-­‐ college students in CA. If we can study the whole popula>on, we do not need to go through the sampling procedures. Much research is based on samples of people. Representa5veness -­‐ how representa5ve the selected cases are? Then, can knowledge gained from selected cases be considered knowledge about a whole group of people? The answer depends on whether those selected cases are representa(ve of larger group. Newsmagazine ar>cles about public opinion: How can we be sure that the results reflect the public’s true opinion, in other words, how much they can represent views of all Americans. The ul>mate purpose of sampling is to get accurate representa(veness. The important considera>on about samples is how representa>ve they are of the popula>on from which we draw them. Casual vs. scien5fic sampling In both daily life and prac>ce, we are involved in sampling decisions -­‐ movies, car purchases, class selec>ons, etc; to get feedbacks about service sa>sfac>on from clients – what is said in community or agency mee>ng. How much of this informa>on is representa>ve? The informa>on can be misleading or biased -­‐ The people who aLend or are the most vocal at a mee>ng may be the most sa>sfied (or most dissa>sfied). If a sample size is too small, informa>on can be biased as well. Scien>fic sampling is considerably more careful and systemic than casual, everyday sampling. In research, scien5fic sampling procedures have been developed so that we can minimize the likelihood that samples we select will be biased or too small.

Sampling Terminology

(Study/Target) Popula5on All possible cases of what you are interested in studying. This is the group you would like to sample from because this is the group you are interested in generalizing to. A sample is drawn from a popula5on. The target (study) popula5on oKen is people who have some par5cular characteris5cs in common, such as all American, all eligible voters, all school-­‐age children, and so on. To select a good sample, we need to clearly define the popula5on from which you draw the sample. The defini5on of popula5on should specify four items: Content: the par>cular characteris>cs that the members of the popula>on have in common Unit: unit of analysis Extent: spa>al or geographic coverage Time: the temporal period during which a unit must possess the appropriate characteris>cs to qualify for the sample

Sampling Terminology (Cont.)

Sampling Frame

-­‐-­‐ A lis5ng of all element (cases) in a study popula5on. -­‐-­‐ In many studies, we draw the actual sample from this lis5ng.

Examples: lis5ngs of telephone numbers, or customers from a local electric u5lity

tend to exclude certain groups, such as….???? telephone books???? case lists????

The adequacy of sampling frame is crucial in determining the quality

of the sample. In other words, the degree to which the sampling frame includes all members of the popula5on is most important. Some of the adequate sampling frames consist of lists of members of organiza5ons. For example, membership rosters of NASW, APA. The popula5on consists of the sampling frame, and we can make legi5mate generaliza5ons only about the sampling frame. Many social workers, for example, do not belong to NASW. Thus, a sample taken from the NASW membership roster represents only NASW members and not all social workers. It is important to assess carefully who the list includes and who the list excludes.

Sampling Techniques

Probability Sampling -­‐ based on probability theory -­‐ equal probability of selec1on can ensure representa>veness -­‐ random selec1on -­‐ Can es>mate sampling error Types of probability sampling Simple Random Sampling Stra>fied Sampling Cluster Sampling

Simple Random Sampling (SRS)

Ø Applies random selec>on using random numbers Ø Generally assumed in all probability sampling applica>ons. Ø Once a sampling frame has been established, the researcher assigns a single number to each member in the list without skipping any number in the process. Ø A table of random numbers is some>mes used to select element for the sample Ø Many computer programs can generate a series of random numbers. Ø SRS is easy to accomplish and explain to others. Ø SRS oUen is imprac>cal, especially not most sta>s>cally efficient when dealing with large scale projects. Ø Simple vs. not most efficient & not good representa>on of subgroups To deal with these issues you may turn to other sampling techniques.

Cluster Sampling

You may have to sample a popula5on that is spread across a wide geographic region. Imagine taking a SRS of all the residents of CA to conduct personal interviews. By the luck of the draw, you will wind up with respondents who come from all over the state. Your interviewers are going to have a lot of traveling to do. Cluster random sampling addresses the problem. Use the follow steps: (1) Divide a popula>on into clusters (usually along geographic boundaries) or some popula5ons are already grouped into clusters (e.g., churches, schools) ↓ (2) randomly select clusters ↓ (3) measure all elements within sampled clusters. We carry out cluster sampling mainly because of efficiency of administra5on, and to reduce the need to even larger samples.

Nonprobability Sampling

We can use probability sampling techniques only when we can have a sampling frame. Many 5mes in social work research, it is impossible to develop a sampling frame of a popula5on. For example, undocumented immigrants are, by defini5on, not listed anywhere Perhaps you just want to evaluate a program in a your agency and don’t care whether the program would work with other people in other places and at other 5mes. Researchers oKen must rely on their judgment to purposively select a sample that seems representa5ve or that seems to fit the purpose of the inquiry. Some5mes they simply rely on available subjects All of the sampling techniques that do not involve the use of probability methods are called nonprobability sampling (do not involve random selec5on). Generally, probability sampling is the preferred method. However, in some circumstances in applied social research, it is not feasible, prac5cal, or theore5cally sensible to use random sampling. When you use nonprobability sampling, it is important to be cau5ous and mindful of the risks inherent in it. (^11)

Quota and Snowball Sampling

Quota Sampling You rely on available subjects but strive to representa5ves by construc5ng matrix represen5ng one or more characteris5cs (gender, age, educa5on, religion, race, etc.), and then collec5ng data from people who had all the characteris5cs in a given parameter. You con5nue sampling for each cell un5l you get the desired number and then stop. If you already have 40 women for your sample, but not the 60 men, you would con5nue to sample men. If eligible women respondents come along, you would not sample them because you have already met your quota. Although using quotas may improve representa5veness, it is s5ll nonprobability sampling and rely on available subjects -­‐ oKen depending on who comes along when. Snowball Sampling You begin by iden5fying people who meet the criteria for inclusion in your study. Then you ask them to recommend others they know who also meet the criteria. Subjects are accumulated gradually in a snowball fashion. For example, if you are studying undocumented immigrants, you are not likely to find good lists of immigrants within a specific geographic area. However, if you iden5fy one or two, you may find that they know who the other illegal immigrants in the area and how you can find them. Useful in inves5ga5ng in sensi5ve topics, such as child abuse or drug use, where the perpetrators or the vic5ms might hesitate to iden5fy themselves if approached by a stranger, such as a researcher, but might be open to an approach by someone who they know shares their experiences or deviant status. Although probability sampling is less risky than nonprobability sampling from the standpoint of generalizing accurately to a popula5on, some good studies use nonprobability sampling. Rather than think you must always use probability sampling, you should understand when certain sampling techniques are more appropriate than others, the func5ons of each techniques, how feasibility constraints bear on the choice of a sample procedures, and the risks inherent when samples are too small or vulnerable to bias.

Extreme Case, Sequential, and Theoretical Sampling — Extreme Case – select cases based on their unusualness or difficulty of finding — Sequential – select cases based on some preset order of selection — Theoretical – select cases according to theory

Sensitivity to Diversity in Sampling Regardless of whether you are using probability or nonprobability sampling techniques, you should be careful to avoid biases: Ø Gender bias – do not generalize to both genders when one gender is not adequately represented in the research sample. Ø Use cultural sensi5vity in all phases of research, including sampling. Ø Carefully examine whether certain minority groups are inadequately represented in the sample, or unwarranted generaliza5ons are made to the en5re popula5on. Ø Employ local community members as research staff Ø Providing transporta5on and child care or data collec5on at home Ø Choose a sensi5ve and accessible sebng -­‐ may not want to travel to strange area, convenient and safe sebng Ø Use and train culturally competent interviewers & bilingual staff