Skip to main content

Table 1 Training area under the receiver operator characteristic curve quantities

From: Statistical learning techniques applied to epidemiology: a simulated case-control comparison study with logistic regression

Method

Az

σ

SE

LR

0.791

0.028

0.008

LRint

0.814

0.027

0.008

k

0.958

0.013

0.004

kx1

0.867

0.023

0.007

kx2

0.728

0.031

0.009

x1

0.490

0.035

0.011

x2

0.772

0.029

0.009

  1. This table gives the area under the receiver operator characteristic curve (Az) quantities derived from the training dataset for the standard logistic regression model with x1 and x2 (LR), the logistic regression model with x1 and x2 with x1 × x2 interaction (LRint), the statistical learning (SL) techniques using a kernel mapping with x1 and x2 simultaneously (k), and partial SL-kernel models using x1 (kx1) and x2 (kx2) individually. This also gives the Az quantities for the x1 and x2 case-control training distribution samples estimated without using model processing. Az and σ are the respective means and standard deviations summarized from the bootstrap trials. SE is the standard error in Az.