I've transformed a few DVs for my regression models and all is cool - the residuals now look nice and normal and homoscedastic etc (before they looked less good because the variables were quite positively skewed). But... now I'm wondering how on earth to interpret the betas.. normally for one unit increase in your IV there is an increase of X amount (beta) in the DV. But how would one interpret this when the betas are based on log and square root transformed variables?

I'm half thinking it is better to just go with the slightly biased non-transformed data. The pattern of results is quite similar, and at least then the interpretation is dead straight forward!

Any advice appreciated!

Just to follow up on my original question (in case anyone encounters it in the future and it is helpful) - I decided to just show the untransformed values in the models because the differences were so minimal and it wasn't worth the trade off of the results being super difficult to interpret vs slightly better accuracy. I have noted simply noted where the differences occurred with the transformed data and what they were.