significant at the .05 level. The dependent variable was the

number of months the patient survived after the diagnosis.

The predictor variables selected for lung cancer were age,

stage, initial treatment, and subsequent treatment. For

breast cancer, they were stage and subsequent treatment.

And, for rectocolon cancer they were sex and stage.

 The results show the efficacy of these variables as

predictors of survival. This is demonstrated by a comparison

of these results to the findings of an earlier study (Worden,

Harrison, & Johnston, 1974). A summary of regression data

from the 1974 investigation and from the present study is

given in Table 33.

 The regression data from the earlier study (Worden,

Harrison, & Johnston, 1974) are a part of the analysis of

data from six cancer sites that was carried out in the

development of this procedure. For two of the three sites,

the regression equations developed for the present study

represent improvements in the explanation of variability in

survival by the effects of biological, disease, and medical

variables. Also, in the prediction for all three sites the

number of variables included in the regression equations is

considerably reduced. A further demonstration of the

efficiency of these equations for all three cancer sites is

that there is a reduction in the standard error of estimate.