Second, we estimated associations between the stressors and each health outcome using an identical series of models. Ordinary least-squares regression was used for poor SRH and depressive symptoms, while Poisson regressions were used for functional limitation and chronic illness. Model 1 estimated the association between each stressor and health. Model 2 estimated the effects of all eight stressors simultaneously. Model 3 included a series of indicator variables for the number of stressors. Model 4 included variables for both the types and number of stressors. In this model, the number of stressors started at exactly two, because the value of the variable for exactly one stressor domain was perfectly predicted by the eight indicator variables for types of stressor domains. All models adjusted for age, gender, race, language of interview, education, and income.
