**Learning Objectives**

- Check model assumptions and fit of the Cox model
- residuals analysis
- log-minus-log plot

- Fit and interpret multivariate Cox models
- perform tests for trend
- predict survival for specific covariate patterns
- predict survival for adjusted coefficients

- Explain stratified analysis
- Identify situations of competing risks
- Describe the application of Propensity Score analysis

- Vittinghoff sections 6.2-6.4

**Outline**

- Review
- Assumptions of Cox PH model
- Tests for trend
- Predictions for specific covariate patterns
- Stratification
- Competing risks
- Propensity Score analysis to control for confounding

**Learning Objectives**

- Make stratified and unstratified Kaplan-Meier plots
- Perform Cox proportional hazards regression
- Assess proportional hazards assumption

**Exercises**

- Load the Primary Biliary Cirrhosis (pbc) dataset from the survival package
- Create a
`Surv`

object using variables “time” and “status”, add this to the pbc dataframe - Plot a KM curve for all participants using
`library(survminer)`

function`ggsurvplot()`

- Stratify by treatment and add a p-value to this plot (see
`?ggsurvplot`

) - Check whether these p-values correspond to results from a log-rank test
- Perform a Cox proportional hazards regression, using the “trt” variable as a predictor
- Create a log-minus-log plot to test the proportional hazards assumption
- Plot Schoenfeld residuals and perform Schoenfeld test for the above Cox model