Causal Analysis in Theory and Practice

September 23, 2009

Differences Induced by Adding Covariates

Filed under: Opinion — moderator @ 3:00 am

Donald Klein writes:

Unfortunately, much of this is over my head. As a practicing trialist the apparent disagreements re managements of # of covariates in ANCOVA seems important but I can’t find a reference to an actual trial where these alternative analyses produced different results. Even a model trial would be helpful to show the import of these theoretical differences.

Similarly with regard to missing data , which turns the experimental randomization towards a naturalistic study.

Cordially,

Don Klein

2 Comments »

  1. Donald,

    I can’t find a reference to an actual trial where these alternative analyses produced different results. Even a model trial would be helpful to show the import of these theoretical differences.

    Differences induced by adding covariates to the analysis are routine matter in every observational study. Otherwise one would not bother adding those covariates. It is normally assumed, however, that the associations measured with more covariates are necessarily less biased. The discussion refutes the wisdom of this assumption.

    The danger of overadjustment is minimal in randomized trials Chapter 11 in Causality shows cases where, even in randomized trial, stratifing on bad covariates produces bias.

    You write:
    Similarly with regard to missing data , which turns the experimental randomization towards a naturalistic study.

    I would like to answer this question, but I do not understand why the problem of missing data in randomized trial is any different that missing data in any other statsitcal estimation exercise, say surveys.

    Judea Pearl

    Comment by judea — September 28, 2009 @ 10:17 pm

  2. Google usually launches one new smartphone under its Nexus brand per year, but this time the search giant might be planning to release two devices, according to a report from Android Police.

    Comment by Christian Dior iPhone 6 Case — May 28, 2015 @ 12:34 am

RSS feed for comments on this post.

Leave a comment

Powered by WordPress