Module_5

  Overview of Sampling When we conduct research, we need to select units of study. Sampling refers to the process of selecting //units of study// (e.g., person, institutions, and books). Sampling is essential for **//generalization//** of current findings to a larger universe that is not within the scope of present study, which means that your study results can be applied to general population if your samples are representing that population. Sampling is important for both quantitative and qualitative research in order to produce //unbiased// findings and have significant //impact// in the field, but generally //less// emphasized in qualitative than quantitative research.

Objectives In this module you will learn to distinguish Target population vs. accessible population Target sample vs. final sample Different sampling methods Random sampling vs. non-random sampling Representative sample

 Population Population is the //universe// or collection of units of study to which the results are to be generalized and from which samples are selected. There are two levels of population. One is the //target// population which is the intended group for generalization justifiable by the current study, and the other is the //accessible// population which refers to the realistic group to which the researcher has potential access (each unit has the potential possibility of being selected). For example, in our research study we decide we want to know the attitude of high school principals towards the No Child Left Behind Act in the U.S. We know it is impossible for us to collect data from **//__all__//** of the high school principals across the county; however, we can access high school principals in Orlando to collect our data. In most cases we can also reasonably expect that the high school principals in Orlando that would be included in our study will have similar attitudes as the majority of the principals in the U.S. //Target population: High school principals in the U.S.// //Accessible population: High school principals in Orlando in 2009// However, if there is reason to believe that Orlando high school principals may not be representative of the target population, then the researcher should designate "high school principals in Orlando" as the target population because the results of this study can only be generalized to the principals in Orlando. Researchers have the responsibility of demonstrating the approximation of the accessible population to the target population.

Sample //Sample// is the subset of the population from which information is collected. Like population, there are two levels of samples. One is the //target// sample which refers the units originally selected for the study. The target sample is usually selected from the accessible population. However, sometimes we may lose some units (participants) from our target sample, such as participants that drop out or disagree to participate in the study, and invalid cases, all the remaining units in our study are called the //final// sample. The following flow chart shows the sampling procedures:

Relationships between Population and Sample To better understand the relationship between population and sampling, please review the " Let's Learn About Sampling " video. You will need the latest version of the Adobe Flash Player to view the video. This can be downloaded at the http://www.adobe.com/shockwave/download/index.cgi?

Sampling methods are generally classified into //**random sampling **// and //**non-random sampling **//.

Random Sampling In random sampling, every unit has an //equal chance // of being selected. Random sampling is always desirable and preferred over nonrandom sampling whenever possible, unless sample size is small where randomization is hard to realize. We should note that random sampling is usually not feasible in qualitative research because of the nature of the intensive engagement in the research settings. Nonrandom sampling is justifiable only if sufficient information about the sample's characteristics are provided so that judgment about generalizability can be made. Simple Random Sampling Simple random sampling selects units from the target population without stratification of the units within the study. When a sample size is small it is difficult to ensure that the sample distribution will be representative of the larger target population. In other words, the sample may not reflect the exactly proportions of characteristics of its population when the sample size is small. Stratified random sampling Stratified random sampling selects units with the consideration of proportions of subgroup characteristics that are representative of its population, such as gender or ethnicity, so it can guarantee sample distribution by strata. The following graph shows the concept of stratified random sampling procedures.

Random cluster sampling Random cluster sampling randomly selects intact groups from target population, such as classrooms, schools, or working units. Multi-stage random sampling Random clusters are selected at a prior stage (//random cluster sampling //), and then random units are selected from those clusters at a later stage (//simple random sampling //). Graphics for Multi-stage random sampling

Non-Random Sampling Systematic sampling Every //n //th member in the list is selected for inclusion in the sample. For example, in class A and class B, the second student on the list in each class is selected. In systematic sampling we need to watch out for the periodicity effect when the population list contains a pattern that coincides with the sampling interval, such as the following table in which the second student in each class has a similar GPA score. Systematic Sampling Example Table

Purposive sampling In the purposive sample, the sample selection is based on personal judgment based on previous knowledge or the belief that the sample has some representative characteristics to serve the specific purpose of the research. In purposive sampling, we need to watch out for //subjectivity //in order not to produce biased study results. Convenience sampling In convenience sampling, samples are selected from any member that comes in handy, such as volunteers or through an arbitrary selection. Convenience sampling is least desirable because it is hard to justify that the sample is representative of the population. In general, purposive sampling and convenience sampling are typically used in qualitative research other that quantitative research because quantitative research emphasizes the generalizability of their research studies which will normally require representative samples.

5.4 Population Generalizability and Sample Size Generalizability Samples relate to or influence the external validity of research studies. Final sample characteristics must be similar to the designated target population in order to generalize the results of research to that target population. In the following table we have samples that contain students reflecting similar proportions of gender, schools, and IQ levels as the population.

The other aspect of external validity is //ecological validity//, which relates to the generalizability of the sample to other settings or conditions. We will discuss more about validity in a separate section.

Determining Sample Size Sample size should be //decreasingly proportional// to population size (i.e., the larger the pop size, the smaller the sampling ratio) The greater the //group heterogeneity//, the larger the sample size The lower the //expected return or retention rate//, the larger the target sample The higher the desired //statistical power// for detecting small differences or statistical precision, the larger the sample size We can conduct a //power analysis// to ensure adequate sample size (generally >= 30) for quantitative analysis We need to keep in mind that //representativeness// takes precedence over //size// (larger is not necessarily better) Qualitative research typically requires a smaller sample size because of its intensive engagement in the research setting Sample cell sizes in stratified sampling may be either fixed (equal) or proportional to population (if fixed, need to adjust by sample weight)

To evaluate sampling procedures, we could ask the following questions: Are the target and accessible population delineated? Is the sampling method appropriate? Is the return rate or retention rate acceptable? Is the final sample size justifiable? Is the final sample demographics approximate target population?

Go to the //Sampling Exercise// //(The video in the first part of the module to help you to review the content and the related exercise)**.**// This will give you the opportunity to determine the different stages of sampling. You will have unlimited attempts to work through the exercise.

If you would like to practice for the mid-term, go to the textbook website to access the Chapter 6 multiple choice and matching activities as a self assessment activity. The link is as follows: []

Please continue working in your groups or individually (if elected not to be in a group) to develop your literature review sections and discuss and formulate your sampling plan section for your research proposal.

Please post a brief description of your sampling plan for your research proposal. Your posting should include: What your target population will be What your accessible population will be Your anticipated target sample size, and Your proposed sampling method.