Stratified Analyses
The purpose of a stratified analysis is to test the hypothesis whether identical regression models are appropriate for different groups, that is, whether the relationships between the independent variables and survival are identical in different groups. If a stratified analysis is requested, Survival Analysis first fits the respective regression model separately within each group. The sum of the log-likelihoods from these analysis represents the log-likelihood of the model with different regression coefficients (and intercepts where appropriate) in different groups. Statistica fits the requested regression model to all data in the usual manner, that is, ignoring group membership, and compute the log-likelihood for the overall fit. The difference between the log-likelihoods is then tested for statistical significance (via the Chi-square statistic).