What is it about?
Shows that one must apply a penalty for non-orthogonality when using hybrid ANCOVA and rank test approaches.
Featured Image
Why is it important?
Adding covariates reduces the residual variance but (to the extent that covariates are not orthogonal to treatment) inflates the multilplier for the variance of the treatment effect. Unless this inflation is allowed for, parametric tests based on ANCOVA rank test hybrids will have an inflated type I error rate.
Read the Original
This page is a summary of: A note on non-parametric ANCOVA for covariate adjustment in randomized clinical trials, Statistics in Medicine, January 2003, Wiley,
DOI: 10.1002/sim.1583.
You can read the full text:
Resources
Contributors
The following have contributed to this page
