The levels of stage and subsequent treatment were coded

as ones and zeros in the analysis. For illustrative pur-

poses, the values of b for the levels of these independent

variables are contained in brackets.

Rectocolon Cancer

 As in the previous cases, the development of a final

prediction equation for rectocolon cancer began with two

exploratory regression analyses; one of a model with main

effects only and one with main and selected interaction

effects. Both analyses were of 156 of the sample of 166 due

to the exclusion of cases with missing information pertaining

to histology and/or stage. Once again, a comparison of the

distribution of values on all variables revealed only slight

differences between the two groups. This comparison is given

in Tables 47-51, Appendix F.

 The first regression model included the main effects of

age, sex, race, stage, histology, and initial and subsequent

treatment. The dependent variable in the analysis was survi-

val. The results of this analysis are reported in Table 17.

 The regression was significant F (20,135) = 2.56, p =
 2
.0007 with R = .28. Based on the significance levels of

the partial regression analyses, two independent variables

were retained for further study. These were stage and sex.