Effect size and multiple linear regression and y values When evaluating the effect size of parameters in a multiple linear regression using standardized variables(via z-scores), is it also necessary to standardize your y-values? Would a transformation of the dependent variable (eg. logarithmic) affect the interpretation of the effect size of your beta values?
 A: The coefficients (betas) from a multiple regression model are already measures of effect size.  In general, there are two kinds of effect sizes: unstandardized (raw) and standardized.  The betas from you model using the data as given constitute unstandardized effect size measures.  Unstandardized measures are best if you can count on your audience being familiar with the units of your data.  For instance, a cariologist will be very familiar with mmHg, but people outside of medicine may not be.  
If you can't rely on your audience understanding the units used, you could use a standardized measure of effect size.  If you only standardized X (or only Y) first, I suppose you could say that was 'partially-standardized', although that's a bit of a neologism.  If I thought the consumers of my study would be familiar with only one of the variables (say, Y), I might only standardize the one I thought they wouldn't already have a handle on.  If I couldn't count on them knowing either, I would standardize both.  That's the logic here.  Note that none of these options is really 'right' or 'wrong'.  

Regarding your last question, yes, a log transform would change the interpretation.  That's because the logarithm is a non-linear transformation, whereas standarization is a linear transformation.  Regarding the interpretation in that case, you should read this CV thread: Interpretation of log transformed predictor.  
