Infoskripsi arrow Reference arrow Theory arrow Sampling Techniques: Stratified Sampling
Sampling Techniques: Stratified Sampling





Stratified Sampling In many educational studies, it is desirable to select a sample in such a wav that the research worker is assured that certain subgroups in the population will be represented in the sample in proportion to their numbers in the population itself. Such samples are usually referred to as stratified samples. Let us say, for example, that we wish to conduct a study to see if there are significant differences on Thematic Apperception Test aggression scores of pupils at different ability levels selected from ability-grouped sixth-grade classrooms. Under this grouping system, pupils are classified into three levels on the basis of general intelligence and placed in classrooms accordingly. In this case, if we were to define the population as all sixth-grade pupils in the district being studied and select a random sample, our random sample may not include a sufficient number of cases from one of the three ability levels. In this research we must also consider the possibility that girls will react differently in terms of aggression scores than boys. In: order to avoid samples that do not include a sufficient number of pupils of each sex at each ability level, a stratified sample can be selected. All sixth-grade pupils in the district would be divided into one of the following six groups: superior boys, superior girls, average boys, average girls, slow boys, and slow girls. Subsamples would then be selected at random from each of the six groups. The proportion of subjects randomly selected from each group usually is the same as the proportion of that group in the target population. Therefore, if slow girls made up 8 percent of the sixth-grade population, they should also make up 8 percent of the sample. If this procedure is not followed, any analysis based on the total sample (i.e., all six groups combined) will produce inaccurate information. Suppose, for example, that we randomly selected 100 pupils from each of our six groups. Any statistics, such as the mean, that we computed on these 600 pupils would not accurately reflect the population since the proportion of average pupils in the population is higher than the proportion of superior or slow pupils, and even within ability levels, the proportion of boys and girls are different. The size of the sample is usually determined by the minimum number of cases we decide is acceptable in the smallest subgroup. If we dedde that the smallest group must contain 30 cases, then we select a total sample large enough so that the correct proportion of our smallest subgroup will equal 30. For example, if 8 percent of our sample must be slow girls and this subsample must be 30 cases, then our total sample would be 375 (i.e., 30 - .08). Stratified samples are particularly appropriate in studies where the research problem requires comparisons between various subgroups. In summary, stratified sampling procedure assures the research worker that the sample will be representative of the population in terms of certain critical factors that have been used as a basis for stratification, and also assures him of adequate cases for subgroup analysis.