back transformation of unstandardized regression coefficients

[ssamdani]

Hi all,

 

I have my dependant variable square root transformed. my question is related  to the interpretation of the unstandardized regression coefficient. Normally, a unit increase in certain IV would imply an ‘unstandardized regression coefficient’ increase or decrease in the DV. in this case, they would correspond to increase or decrease in the square rooted DV rather than the DV itself. how do I back transform the regression coefficients. do i just square them? is it that simple or is there another way or interpreting them.

PS. only the DV is square root transformed in my case. all IVs (and there are quite a few of them) are in their original form 

Regards

Sarah

[ssamdani]

Hi Dave. thanks for the input. my skewness values are ok. they are below 1. I am using a criteria +/-2 for both kurtosis and skewness. my main issue is kurtosis. I did try the log transform but t does not help me with bringing down the value for kurtosis. a square root transfor does that for me. I hope I am making sense. can an alternative way be using the standardized coefficients of regression. I also noticed that transforming the DV although brings down the kurtosis value but does not improve R square any further. in that case do you think it makes sense to report the untransformed resutls.

thanks in advance.

[Author]

Posts