Skip to main content

Table 2 Variable selection using nested LRT in Test set

From: Prognostic significance of cytoplasmic S100A2 overexpression in oral cancer patients

Fitted Cox models p*
Nuclear vs. Cytoplasmic S100A2 attributes
Nuclear variables 0.70
Cytoplasmic variables <0.01
Combined nuclear and cytoplasmic variables 0.04
Nested LRT 0.34
Most predictive subcellular variables Cytoplasmic S100A2
% Positivity alone vs. % Positivity & Intensity
Cytoplasmic variables <0.01
Without cytoplasmic intensity <0.01
Nested LRT 0.07
Most predictive variable cytoplasmic % positivity
  1. *p corresponds to the LRT of the model fit, unless stated otherwise.