BSTA 6652 Survival Analysis Winter, 2017 Problem Set 6

We will analyze the dataset valung_small.csv (on Blackboard, under “Course Materials”), which contains 48 observations and 7 variables, described as follows:

Name Description therapy Type of Treatment 0:standard 1:test time Failure or Censoring Time in Days dead 1:dead 0:censored kps Karnofsky Performance Status 10-99 diagtime Months From Diagnosis to Entry age Age in Years prior Prior Treatment 0:no 1:yes

The data were collected before 1980 to assess the effectiveness of the testing therapy on lung patients, compared to the standard therapy. INCLUDE SAS CODE FOR EACH PART.

(a) Fit the Cox proportional hazards model containing all covariates using the Breslow, Efron and exact methods for handling tied survival times. Show the output of parameter estimates. Compare the estimates of the coefficients and standard errors from the 3 methods. Which method would you recommend and why? (b) Using the method you pick and the (univariate) Wald test, which variables appear not to contribute to the model fit at the 5% significance level? Remove all such covariates and fit the reduced model. Also test for the adequacy of the reduced model (compared to the full model) using the likelihood ratio test. (c) Conduct Backwards Model Selection. Is the final model the same as the reduced model in part (b)? Test for the adequacy of the final model using the likelihood ratio test. (d) Fit your final model. What is the fitted final model? Please do not print out the baseline survival estimates for all time points. Instead, draw the baseline cumulative hazard function. Interpret the effect(s) of significant covariate(s). For insignificant covariates, you can just state the covariate is not significant. (e) Consider two patients both at age of 60, having kps of 50, no prior treatment, and 4 months from diagnosis to entry, but one was treated by the standard therapy while the other by the test therapy. Draw their predicted survival curves. Do they look significantly different to you?

1