### Instructor

John E. Kolassa, kolassa@stat.rutgers.edu

### Announcements

• Office hours are 5:30-6:30 Thursday, and 2-3 Wednesday.
• Midterm exam solutions here.

### Course information

Course Information Schedule with Readings from Main Text Schedule with Readings from Main and Recommended Texts Student Survey An old final

### Lectures

• Lecture 1 (small format)
• Lecture 2 (small format) (SAS input) (SAS output) (graphs) (R code) , plus the larynx data and the weening data.
• Lecture 3 (small format) (SAS input) (SAS macro) (SAS output) (graphs) (R code) (R function to do actuarial estimate) (R code to simulate simultaneous coverage)
• Lecture 4 (small format) (SAS input) (SAS macro) (SAS output) (R code) (R function to do Renyi and Cramer-von Mises testing) (graphs)
• Lecture 5 (small format) (examples) (SAS output) Schoenfeld paper (Critical values for the Renyi statistic)
• Lecture 6 (small format) (examples) (SAS output) (R code)
• Lecture 7 (small format) (examples) (SAS output) (R code) (graphs)
• Lecture 8 (small format) (examples) (graphs) (R code for some of the examples)
• Lecture 9 (small format) (examples) (SAS macros) (R code) (R macros) (graphic output)
• Lecture 10 (small format) (examples) (SAS output) (graphic output)
• Lecture 11 (small format) (examples) (SAS output) (graphs) (macros) (R code) (R macros)
• Lecture 12 (small format) (examples) (SAS output)
• Lecture 13 (examples) (graphs)
• Review (examples) (SAS output) (graphs)

### Prerequisites

Level V statistics is presumed. This prerequisite may be met by 563 or 586 or 593.

### Other resources

• Software:
• SAS documentation online. SAS versions 8 and 9 give different results for some calculations I offer in class. Specifically, Version 9 allows SAS to calculate internally transformed confidence intervals for survival curves. In class I did this using PROC PHREG. Unfortunately, the default that Version 9 uses is different from what you get from Version 8. Specifically, Version 9 uses as the default the complementary log-log transform, while Version 8 used no transformation. Hence if the data set mydata contains two variables, t a time variable and s a status variable, taking the values 0 for censored and 1 for event, then the code
``` proc lifetest data=mydata outsurv=a; time t*s(0); run; ``` puts into a confidence intervals on the original scale in version 8, but on the log log scale in version 9. To get confidence intervals on the original scale in version 8,do
``` proc lifetest data=mydata; survival out=a conftype=linear; time t*s(0); run; ```
• The R program. You may have to separately download and install the survival analysis library. Some of the material we will use is in the surv2sample package.
• If you have a hard time reading the pdf files I put up, and are using Windows, you might try downloading gs5.50, a postscript reader that will also read pdf, or Acrobat Reader
• Data: