Modern Survival Analysis
Dr. Terry Therneau, Mayo Clinic, Rochester, MN
The Section of Biostatistics at the Mayo Clinic in Rochester, MN provides statistical analysis services for research work in nearly all Mayo departments. They have archived results on over 2,000 studies since 1970. With its services in such demand, the Biostatistics Section places a premium on high-quality, efficient data analysis. According to Dr. Terry Therneau, Head of the Section of Biostatistics, the best data analysis solution available today is not one product, but two. Therneau and his colleagues use both S- PLUS and SAS to get the most out of their data analyses.
For the February 7, 2000, Short Course in Indianapolis, Dr. Therneau will select topics from the following outline for the 2-day course that he has taught several times.
1. Preliminaries
Notation and Conventions to Be Used
Five Minute Introductions to S-Plus and SAS (enough to read the overheads)
Definition of the Cox model
The Counting Process Form of a Cox Model
Mathematics
History
Particular examples
Time dependent covariates
Discontinuous intervals of risk
Multiple time scales
Implementation in SAS and S
Examples
Stanford heart transplant
Parkinson's disease
2. Leverage and Robust Variance
Leverage
Jackknife
Approximate jackknife methods
Example: PBC data
Robust Variance Estimation
Jackknife
Grouped jackknife
Relationship to sandwich estimates and other methods
3. Multiple and Correlated Events
Methods
Time to first event
Special time-dependent covariates
Random effects models
Marginal models
Motivation for the Marginal Model
Simple Examples
Doubled data
Delayed second event
Model choices
Ordered events vs unordered events
Time scales
Strata Time-dependent covariates
Counting process form
Models
Andersen-Gill
Marginal
Conditional
Particular cases
Different types of event
PBC UDCA study
Multiple occurrence
rhDNase study
Gamma interferon
Bladder cancer study
Paired events
Diabetic retinopathy study
Survival and progression in colon cancer
Complex Examples
Homeless and Shelter
Multiple States in Inflamatory Bowel Disease
Random Effects or "Frailty" Models
Relationship to penalized models
Gamma random effects
Gaussian random effects
Degrees of freedom
Examples
Rats
Colon cancer
Gamma interferon study
4. Summary and Recommendations