Comments on: Stats and Methods Urban Legend 2: Control Variables Improve Your Study http://neoacademic.com/2011/04/27/stats-and-methods-urban-legend-2-control-variables-improve-your-study/ technology, education and training research from an industrial/organizational (I/O) psychologist in the ivory tower Sat, 04 May 2013 15:02:58 +0000 hourly 1 http://wordpress.org/?v=3.5.1 By: Jeromy Anglim http://neoacademic.com/2011/04/27/stats-and-methods-urban-legend-2-control-variables-improve-your-study/#comment-8033 Jeromy Anglim Wed, 27 Apr 2011 14:37:20 +0000 http://neoacademic.com/?p=1230#comment-8033 I’ve just been reading this special issue of ORM. It’s a good one. Thanks for summarising the article on covariates for the blogosphere.

I often encounter this problem of inappropriate use of covariates.
I’d add a few other issues:
* not justifying causal mechanisms in a mediational models
* not justifying causal mechanisms in a moderating hypothesis, and
* not justifying causal claims for the direction of a bivariate relationship in an observational study.

A common thread across all these issues, including the covariate one, seems to be an inadequate reflection of causal mechanism.

]]>