Introduction
A multiple logistic regression is a useful statistical tool because it allows for controlling for variables (other than the independent variable) which may influence an outcome. However, it cannot be applied to every situation. In this paper, the topic and research question of the dissertation of the author of this paper are provided, and it is considered whether a multiple logistic regression can be used for this study.
Considerations While Selecting a Methodology for the Dissertation
The general topic of the dissertation of the author of this paper is “A study of a relationship between HIV treatment compliance and social support among African American women with HIV,” and a quantitative methodology will be used to conduct the research. While selecting an exact method to be utilized, it is needed to formulate the research question and consider which method will suit it best.
In this case, the research question that will be addressed is as follows: “What is the influence of perceived social support on HIV treatment compliance among immigrant and non-immigrant HIV-infected African American women?” Therefore, it is needed to compare the influence of the social support on the HIV treatment adherence separately for each of the two groups (immigrant and non-immigrant), and then compare the results.
It might be possible to use two simple linear regressions for this purpose, for they will permit for forecasting changes in the dependent variable for each unit of change in an independent variable. The transcript, however, notes that a logistic regression allows for controlling for other factors which might influence the outcome. Nevertheless, a logistic regression cannot be used in this case, for the outcome variable will not be categorical (Forthofer, Lee, & Hernandez, 2007). Furthermore, a simple linear regression can assess the amount of variance explained by the independent variable (Laerd Statistics, 2013), so it appears that controlling for other factors is not necessary in this case.
Applying a Multiple Logistic Regression to the Given Research Topic
As it has already been noted, the reason why a logistic regression cannot be used for this study is that the dependent variable is quantitative (Warner, 2013); it measures what percentage of the required medication was taken. It might be possible to use a logistic regression if the outcome was categorical – for instance, if it was measured as “adherent to medication” and “non-adherent to medication.” There is, however, a problem in defining these categories, for it appears unlikely that many HIV-infected persons are 100% adherent, and e.g. 96% compliance still appears to be a good rate of adherence.
In this case, a logistic regression could be used to predict this outcome (adherent / non-adherent) from the level of perceived social support. However, it would also be recommended to control for a number of other factors which may influence this outcome, such as the levels of education or health literacy.
Conclusion
Therefore, a logistic regression cannot be used for the given research question because the outcome variable is not categorical. Thus, two simple linear regressions might be used to analyze data pertaining to this problem, and their results may then be compared.
References
Forthofer, R. N., Lee, E. S., & Hernandez, M. (2007). Biostatistics: A guide to design, analysis, and discovery (2nd ed.). Burlington, MA: Elsevier Academic Press.
Laerd Statistics. (2013). Multiple regression analysis using SPSS Statistics. Web.
Warner, R. M. (2013). Applied statistics: From bivariate through multivariate techniques (2nd ed.). Thousand Oaks, CA: SAGE Publications.