Letters Section Editor: Jody W. Zylke, MD, Senior Editor.
Author Affiliations: University of Michigan, Ann Arbor (firstname.lastname@example.org).
In Reply: We observed a substantial gender difference in salary within a relatively homogenous population of physician researchers both in a full model (including 20 independent variables that might confound or mediate the association between gender and salary) and a parsimonious model constructed based on improvement in the Akaike information criterion.
Model selection faces inherent tension between inclusion of adequate variables to ensure goodness of fit and appropriate exclusion to ensure parsimony and avoid multicolinearity. We deliberately included the full model in our article to allay concerns that omission of variables was underlying the gender difference observed. Male gender was significantly associated with salary (+$12 001, P = .006) even when all theoretically selected variables (including work hours, which includes clinical hours and percentage of time in research, which is the main activity pursued by this cohort other than clinical work) were included. Dr Nomura objects to the lack of inclusion of work hours in our second, parsimonious model. If we add work hours to the parsimonious model, gender continues to be associated with salary, with similar effect size and significance (+$12 324, P = .003).
Jagsi R, Griffith K. Salary Differences by Gender—Reply. JAMA. 2012;308(12):1207-1208. doi:10.1001/2012.jama.11177