The Topic of the Research and the Statistical Test to Be Used
The general topic of the research is “A study of the relationship between HIV treatment compliance and social support among African American women with HIV.” More specifically, the influence of perceived social support on HIV treatment adherence among immigrant and non-immigrant African American females will be tested. The statistical tests that will be used are two simple linear regressions (George & Mallery, 2016); the results of these regressions for immigrants and non-immigrants will be compared.
Additional or Alternative Tests in the Study
Poisson Regression
A Poisson regression could be used if the dependent variable consisted of count data (Forthofer, Lee, & Hernandez, 2007). In the proposed study, the dependent variable is a percentage, i.e. it measures what percentage of pills respondents took relative to the number of pills that was prescribed. However, if the study counted how many days the participants took their pills during a certain period of time, a Poisson regression (or, rather, two Poisson regressions, one for each group – immigrant and non-immigrant) could have been used.
To incorporate the test into the study design, the data about HIV treatment compliance would need to be gathered as counts data. The Poisson regressions would be used instead of the simple linear regressions.
The results would be similar to those gained via two simple linear regressions, but the outcome (compliance) would be measured less precisely, apparently making the analysis less exact as well.
Ordered Logistic Regression
It would be possible to utilize an ordered logistic regression if the dependent variable was measured on an ordinal level (Warner, 2013). This could be done if the HIV treatment compliance was assessed using a Likert scale, e.g., a five-point Likert scale. Two ordered logistic regressions would be needed: for immigrants and for non-immigrants.
To incorporate the test into the study design, it would be needed to convert the data about HIV treatment adherence into a Likert scale. Then, the ordered logistic regressions would be used instead of the simple linear regressions.
The results would be similar to those gained via two simple linear regressions, but the adherence would be measured less precisely, so the results of the analysis would also be less exact than those gained with simple linear regressions.
ANCOVA
An ANCOVA could be performed on the same data as that which will be gathered for the simple linear regressions. The groups would be “immigrant” and “non-immigrant,” the covariate would be “social support” (continuous scale) and the dependent variable would be “HIV treatment adherence” (continuous scale).
To incorporate the test into the study design, no adjustments would be needed. In fact, an ANCOVA could be run as an additional test.
The results of the ANCOVA would demonstrate whether there is a statistically significant difference in HIV treatment adherence between the immigrant and non-immigrant Black females while controlling for social support (Field, 2013), and estimate the magnitude of this difference (e.g., the mean difference, and the respective confidence intervals) (Ellis, 2010). On the contrary, the simple linear regressions would show how much the independent variable (social support) affects treatment compliance in the two different groups.
Therefore, for the given study (that measures the relationship between HIV treatment compliance and social support), two simple linear regressions are more appropriate. However, an ANCOVA could be used as an additional test to check whether there is a difference between the two groups.
Conclusion
Therefore, Poisson regressions and ordered logistic regressions could be used instead of simple linear regressions in this study, but the results would be less exact, for the dependent variable would be measured less precisely; information would be lost. An ANCOVA, however, might be used as an additional test for this study without needing to make any adjustments to the data (apart from the possible transformation required if the data violated the assumptions of the ANCOVA).
References
Ellis, P. D. (2010). The essential guide to effect sizes: Statistical power, meta-analysis, and the interpretation of research results. Cambridge, NY: Cambridge University Press.
Field, A. (2013). Discovering statistics using IBM SPSS Statistics (4th ed.). Thousand Oaks, CA: SAGE Publications.
Forthofer, R. N., Lee, E. S., & Hernandez, M. (2007). Biostatistics: A guide to design, analysis, and discovery (2nd ed.). Burlington, MA: Elsevier Academic Press.
George, D., & Mallery, P. (2016). IBM SPSS Statistics 23 step by step: A simple guide and reference (14th ed.). New York, NY: Routledge.
Warner, R. M. (2013). Applied statistics: From bivariate through multivariate techniques (2nd ed.). Thousand Oaks, CA: SAGE Publications.