Multivariate Regressions Predicting Self-Report of Health Problems among Categories with More Than One Correlate
| Variance Accounted For (%) | β* | P | |
|---|---|---|---|
| Organic/major problems | |||
| Survivors (n = 156) | |||
| Total model | 22 | .00 | |
| Age† | 0.15 | .04 | |
| Sex‡ | 0.14 | .07 | |
| Treatment intensity† | 0.38 | .00 | |
| Controls (n = 138) | |||
| Total model | 9 | .00 | |
| Minority status‡ | −0.12 | .03 | |
| Sex‡ | 0.18 | .00 | |
| Constitutional/other problems | |||
| Survivors (n = 156) | |||
| Step 1§ | 11 | .00 | |
| Age | 0.21 | .01 | |
| Treatment intensity† | 0.25 | .00 | |
| Step 2‖ | 7 | .00 | |
| Lymphoma | 0.13 | .13 | |
| Solid tumor | 0.27 | .00 |
* β is standardized.
† Age and treatment intensity are entered as continuous variables to facilitate easier interpretation and to maximize variance of these variables. Thus, positive β values indicate that problems increase with higher age and higher treatment intensity.
‡ Dichotomous variables are coded as 0 and 1. In particular, sex is coded as 0 = male, 1 = female and minority status is coded as 0 = white and 1 = minority. Thus, a positive β value for sex indicates a positive relationship between female sex and problem reporting. The negative β value for minority status indicates a positive relationship between being white and report of problems.
§ Two steps were used in the regression predicting Constitutional/Other problems in order to determine the unique contribution of diagnosis given the multiple categories in this variable.
‖ Dummy codes for diagnosis were used to reflect categories of diagnoses of lymphoma and solid tumor.