User Tools

Site Tools


Prediction models

Stepwise regression

The prediction model was fitted using logistic regression with stepwise regression and backward elimination. Why? Did the exclusion of variables from the model improve the predictive accuracy? A LASSO regression, LAR regression or Elastic Net may perhaps lead to a model with better predictive accuracy.

Interpretation of regression coefficients

The presented regression coefficients of the prediction model can be misinterpreted as estimates of relative risk. Since the model has not been developed with respect to cause-effect relations the parameter estimates are likely to reflect confounding bias and perhaps adjustment bias, see e.g. Westreich D, Greenland S. The table 2 fallacy: presenting and interpreting confounder and modifier coefficients. Am J Epidemiol 2013;177:292-298. I recommend removing these tables from the manuscript.

predicts.txt · Last modified: 2020/02/16 18:55 by ranstam