Feigl and Zelen (1965), Cook and Weisberg (1982), and Johnson (1985) give data on survival of 33 leukemia patients as a function of their white blood cell count and the existence of a certain morphological characteristic in the cells. The characteristic is referred to as either AG positive or AG negative. The binary response is survival of at least 52 weeks beyond the time of diagnosis. The data are given in Table 4.18. Fit a logistic regression model with separate slopes and intercepts for AG positives and negatives. Examine the data for influential observations. Consider whether a log transformation of the white blood cell count is useful. Evaluate models with

(a) the same slope for both AG groups,

(b) the same intercept for both AG groups, and

(c) the same slope and intercept. Examine each model for influential observations.
TABLE 4.18. Data on Leukemia Survival Cell Cell Survival Count AG Survival Count AG 1 2,300 + 1 4,400 −
1 750 + 1 3,000 −
1 4,300 + 0 4,000 −
1 2,600 + 0 1,500 −
0 6,000 + 0 9,000 −
1 10,500 + 0 5,300 −
1 10,000 + 0 10,000 −
0 17,000 + 0 19,000 −
0 5,400 + 0 27,000 −
1 7,000 + 0 28,000 −
1 9,400 + 0 31,000 −
0 32,000 + 0 26,000 −
0 35,000 + 0 21,000 −
0 52,000 + 0 79,000 −
0 100,000 + 0 100,000 −
0 100,000 + 0 100,000 −
1 100,000 +