Hi,

Yes, an effect only becoming significant after correcting for some potentially-confounding factor is a common scenario. One typical example is the effect of medication dosage on health measures, which can often only become significant (and even change direction) when controlling for patient severity (because patients with higher disease severity also typically receive higher dosages of medication; so, when not properly controlled for disease severity, the effect of medication dosage can seem to worsen symptoms, while, when properly controlling for patient severity, medication dosage can be correctly measured to improve symptoms).

Hope this helps

Alfonso

Originally posted by bioash:

Hello Alfonso,

 

I have a question regarding results being valid or not. My connectivity analysis in conn, for any contrast; A > B does not yield any results. However when I add covariates, this give me significant result. In another case the direction of the result changes. For example, after adding covariates the result that was shown for example as increased activation, now shows decreased activation.  

My question is are these resuts valid? Covariate needs to be checked only if original contrast A>B shows significant  results?  What about change in direction of activations when covarites are added? Are the results valid if it does not appear significant for the original contrast, but shows significance after covariates are addd?

 

Thank you