4 ms·
Is there a reason to consider the use of a noisy signal as a strong sign of a low-quality study, without addressing whether the noise is correlated with the out
by boloust 4y ago
Is there a reason to consider the use of a noisy signal as a strong sign of a low-quality study, without addressing whether the noise is correlated with the outcome?
Do we think this study would be improved if it did not control for genetics at all?
- dash2 4y agoIt wouldn't be any worse. The question isn't whether the noise is correlated with the outcome. The question is whether the denoised variable would be much more correlated with the outcome (yes), and whether that would change the effect of the focal variable. In any case, it's a weird thing to control for in a panel study. Why not just use per-person fixed effects? That would eliminate all effects that are constant across individuals.