StrataVariables

is an option for fitting functions such as CoxModelFit that specify the variables on which the model should be stratified.

Details

Strata variables are used to divide a non-homogeneous population into homogeneous subgroups. Stratification is often useful in proportional hazards modeling where the assumption of proportional hazards between subgroups may be suspect.
Possible settings for StrataVariables include:
None treat no variables as strata

{v₁,v₂,…} treat the variables v_i as strata
Models with a strata variable v with p levels {s₁,…,s_p} effectively group the input data into p groups corresponding to the s_i.
With multiple strata variables v₁, …, v_k with p_{v_i} levels, the input data is grouped into at most p_{v_i} groups, dropping empty subgroups.

Define some data:

Fit the data, treating the second variable as a strata variable:

One stratum is created for each level in the data:

A parameter summary table:

Fit a model with more than one strata variable:

All non-empty combinations of the strata variables are used:

Baseline survival rates for the strata:

Top

	None	treat no variables as strata
	{v₁,v₂,…}	treat the variables v_i as strata