The Sampling Strategy for the Research Proposal
The sampling strategy that will be used in the study is simple random sampling (Frankfort-Nachmias & Nachmias, 2015). HIV-infected African American women from the urban setting will be recruited. The procedure of finding the participants will involve contacting some HIV treatment centers and offering its African American female patients chosen randomly from the lists of all such patients to participate in the study. The patients who agree will be asked to complete the survey designed for the study.
This strategy will be appropriate because it is likely that the sample will be representative of the general population, for any HIV-infected persons are likely to use the same local HIV treatment centers. Furthermore, it might be hypothesized that the subgroups of African American females with HIV will be represented approximately proportionally in these centers.
The Strategies that Have Not Been Chosen
This section provides the rationale for not choosing any other sampling techniques described by Frankfort-Nachmias and Nachmias (2015).
Stratified Sampling
The method of stratified sampling might be appropriate for this study, but it seems somewhat redundant. Indeed, there might be different subgroups in the population of African American women who are infected with HIV. However, it appears likely that all of them will attend the same HIV treatment centers, and thus the subgroups will be represented roughly proportionally in a randomly selected sample even without stratifying.
Systematic Sampling
Systematic sampling might also be appropriate for this research. In fact, it seems that it would not be “better” or “worse” than the chosen method of random sampling. The procedure of selecting participants would differ: in systematic sampling, the participants would be chosen from the list of all potential participants by using the sampling interval, whereas in random sampling, e.g., the SPSS randomization procedure (assign random numbers to participants and divide them into groups) can be used. The resulting sample, however, should be random in both cases.
Cluster Random Sampling
The focus of this research is the African American female population with HIV in general, rather than some “natural groups” that may exist in this population. Therefore, cluster sampling is apparently less appropriate for this study than random sampling.
Multi-Stage Random Sampling
Multistage sampling can be treated as a more complicated subtype of cluster sampling (Agresti & Finlay, 2008). It has already been shown that cluster sampling is less suitable for this study than random sampling; therefore, a more complicated form of cluster sampling would also be inappropriate, and would only cause additional complications.
Non-Probability Sampling
Using non-probability sampling would not be appropriate for this study, because this research is aimed at making inferences about the more general population of African American women with HIV, whereas non-probability sampling often does not allow for making inferences about the general population.
Quota Sampling
This type of sampling requires selecting participants according to some known characteristics so that the number of participants in each group would be proportionate to the size of the respective subgroups of the population. However, membership in the subgroups of the population is not the focus of this study.
Noteworthy, this type of sampling might be appropriate for a similar study (for instance, one in which the social status of participants is accounted for). In addition, it may be possible to hypothesize that the population in question is rather homogenous.
The G*Power Analysis
To calculate the desired sample size, G*Power software was used (Buchner, Erdfelder, Faul, & Lang, n.d.). The F-test family was chosen, and the option “Linear multiple regression: Fixed model, R2 deviation from zero” was selected, as advised (Faul, Erdfelder, Buchner, & Lang, 2009). For the current study, α=.05 and the statistical power of.80 will be used; these are the standard levels recommended for use in social research (Trochim, 2006; Warner, 2013). The required sample size to detect a medium effect (f2=.15) with one predictor used is N=55. Of course, sample size can be increased to provide higher power and to allow for detecting a smaller effect.
References
Agresti, A., & Finlay, B. (2008). Statistical options for the social sciences (4th ed.). Upper Saddle River, NJ: Prentice Hall.
Buchner, A., Erdfelder, E., Faul, F., & Lang, A.-G. (n.d.). G*Power: Statistical power analyses for Windows and Mac. Web.
Faul, F., Erdfelder, E., Buchner, A., & Lang, A.-G. (2009). Statistical power analyses using G*Power 3.1: Tests for correlation and regression analyses. Behavior Research Methods, 41, 1149-1160.
Frankfort-Nachmias, C., & Nachmias, D. (2015). Research methods in the social sciences (8th ed.). New York, NY: Worth.
Trochim, W. M. K. (2006). Statistical power. Web.
Warner, R. M. (2013). Applied statistics: From bivariate through multivariate techniques (2nd ed.). Thousand Oaks, CA: SAGE Publications.