Assignment:
The data set wbca contains data from a study of breast cancer in Wisconsin. Read the help information for details.
(a) Fit a logistic regression, with Class as response, and the other nine variables as predictors. Produce a summary of your results.
(b) Would it be appropriate to perform the lack-of-fit test based on the deviance, in this case? Explain briefly.
(c) Form an approximate 95% confidence interval for the probability the tumor is malignant when the predictor values are as follows: Adhes=1, BNucl=1, Chrom=2, Epith=2, Hitos=1, NNucl=1, Thick=2, UShap=1, USize=1. (NOTICE: The variable Class is 0 if malignant, 1 if benign! So the "event" is benign, not malignant.)
(d) Consider classifying a cancer as benign if the probability from the model exceeds 0.5 (and malignant otherwise). Compute the fraction of observations that would be misclassified, under this procedure.
(e) Using the function step, reduce the model. Use direction="backward". (Note that this uses AIC to determine the final model.) Produce a summary of your final model.