The Reviewer’s Guide to Quantitative Methods in the Social Sciences by Gregory R. Hancock & Laura M. Stapleton & Ralph O. Mueller

The Reviewer’s Guide to Quantitative Methods in the Social Sciences by Gregory R. Hancock & Laura M. Stapleton & Ralph O. Mueller

Author:Gregory R. Hancock & Laura M. Stapleton & Ralph O. Mueller
Language: eng
Format: epub
Publisher: Taylor & Francis
Published: 2018-10-23T00:00:00+00:00


2.3. Types of Variables

The types of variables should be appropriate for their respective roles in the model. Classic regression-based moderation involves a continuous IV, either a continuous or categorical ModV, and a continuous DV. Choosing the right type of variables for these three slots causes confusion for some researchers. In particular, researchers want to know whether the IV can be categorical. If the IV is dichotomous categorical (e.g., gender or experimental conditions), then it is more typical (although statistically equivalent) to run the statistical analysis as an analysis of variance (ANOVA). Accordingly, one may occasionally encounter the situation of a dichotomous categorical IV, a continuous ModV, and a continuous DV. In these cases, some researchers choose perform a median-split on the ModV in order to make the variable amenable for ANOVA analysis; doing so, however, can come with a potential reduction in statistical power as well as diminished potential for interpreting the nature of the moderation. Categorical DVs require logistic analyses, and although efforts have been made to systematize so-called logistic moderation (Gelman & Hill, 2007; Hayes & Matthes, 2009), this approach is less frequently attempted in the social and behavioral sciences.

Most categorical ModVs are dichotomous (e.g., asthmatic vs. healthy), although Aiken and West (1991) have shown how ModVs that are composed of multiple categories can be dummy-coded and used in a moderation analysis. For example, ethnic group membership may be constituted by four categories: European American, African American, Hispanic American, and Other. Ethnicity would be dummy-coded, yielding three dummy variables, using one group (e.g., European American) as the reference group. These three dummy variables would be multiplied individually with the IV, and then entered in the regression analysis: the IV, the three dummy codes, and the three interaction terms. Statistically significant interaction terms (one, two, or three) would be graphed individually to enable interpretation.



Download



Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.