Statistics 227

Homework 8

Due Wednesday, 27 November 1996

1. Selvin, Exercise 2, page 407. (The comparison with discriminant analysis is not necessary.)

2. Use the chds (Child Health and Development Study) data to develop a predictive model for the risk of a low-birthweight outcome of pregnancy.

   Create a new variable, lbwt, to have the value one or zero depending on whether birthweight is less than 5.5 pounds (2500 grams). This can be done with the Stata command
   

   
   generate lbwt = (bwt <= 5.5)
   
   [For a description of the chds dataset, see pp. 57-59 of Selvin. Also note that there is a misprint on page 59; the variable fnocig has an extra space inserted in the infile and summarize commands on that page. The space must be removed in order for the commands to work as displayed.]

3. Table 11.17 in Selvin gives data on 35 individuals with high cholesterol levels. Of interest here are factors that predict the occurrence of a heart attack. In the data set, the column labeled "Censored" indicates individuals who had a heart attack (1) and those who did not (0) during the study period. Treating heart attack as the outcome variable, use logistic regression methods to assess the risks associated with weight, cholesterol level, and cigarette usage.

   Write a short paragraph that describes your findings. Do not use the words "significant" or "p-value" in this paragraph.

   [Note that the "Time" variable cannot be used as a predictor. Why not?]