Standardizing b

#1
Hi all,

I'm trying to figure out how to get to the standardized beta in a few papers, because I want to interpret the effect sizes.

Is it correct that for a linear regression this is the way to calculate it?: multiply b with (SD of b / SD y) = beta

And what is the formula for a logistic regression?

Thanks!
 

hlsmith

Less is more. Stay pure. Stay poor.
#2
I can't recall the formula, but in my mind the estimate is standardized by the SD - but I could be wrong. Also, if I recall back, I don't believe the standardized coefficients in logistic reg are quite the same. The concept doesn't carry over completely since the DV is binary.
 

noetsi

No cake for spunky
#3
It has been a long time since I worked with this, but I think there is no agreed on way how to standardize a logistic coefficient. There are many ways to do it. Note that even with linear regression there are well known authors who strongly oppose the use of standardized coefficients (although I no longer remember why). :(