Appendix Table A3.
Incidence Rate Ratios from Poisson Regression Models Predicting the Number of Confidants Lost and Added between Waves (N = 2,140)
| Predictor | Number lost | Number added |
|---|---|---|
| W1 functional impairment (logged) | 1.114 (.083) | 1.033 (.120) |
| W1 self-rated health (reference = “poor”) | ||
| Fair | .954 (.076) | .953 (.075) |
| Good | .988 (.079) | .997 (.093) |
| Very good | .967 (.088) | .936 (.094) |
| Excellent | .982 (.095) | 1.040 (.104) |
| W1 depressive symptoms (logged) | 1.067 (.080) | 1.030 (.094) |
| Change in (logged) functional impairment | 1.103) (.086) | (.088) (.098) |
| Change in self-rated health | 1.000 (.019) | 1.018 (.026) |
| Change in (logged) depressive symptoms | 1.016 (.053) | .936 (.065) |
| Pseudo R2 b | .469 | .085 |
†
p < .10
*
p < .05,
**
p < .01,
***
p < .001 (two-sided tests)
a
Estimates are weighted using NSHAP W1 person-weights (adjusted for attrition and selection at W2).
All models are survey-adjusted and include all of the same covariates and controls that were employed in the analyses presented in Tables 2–4, except for other measures of network change.
b
Represents the squared correlation between respondents’ observed number of confidants lost/added and the corresponding predicted values.