I am reading a study and they find no effect of intervention group on their outcome of WM integrity. They then do a partial correlation (controlling for age, gender, attendance) to see if there is a relationship between change in white matter and change in fitness and they find a signficant correlation. Is this partial correlation a measure of causation? or is it possible that a 3rd variable (other than the ones controlled for) is mediating the relationship? Thanks
1 Answer
$\begingroup$
$\endgroup$
No, a third variable is not a measure of causation. The idea is simply that the third variables might be correlated with both, WM integrity and the intervention measure. If this is the case, and you do not include third variables, then there will be an omitted-variable bias.
It is however still not ensured that the intervention is now, after adding third variables, measured unbiased because there might be other issues involved (wrong specification, still omitted-variable bias, measurement error to name a few). These issues can also mean that adding third variables might introduce an additional bias.
answered Jan 17, 2021 at 15:32