It rather depends on what the variables are. If they are measures of
psychological state, like depression, then the scale units don't
necessarily make any sense anyway, so you just interpret the
standardized estimates.
If they are 'real' units, like money, then the easiest thing to do is
calculate the predicted value of the outcome, at certain levels of the
predictors, and the back-transform to get expected values.
jeremy
On 06/06/07, Pat Burton <[log in to unmask]> wrote:
>
> Dear All,
>
> I'm e-mailing to ask agin for advice on interpretation of regression
> coefficients with transformed data. I have log transformed my dependent
> variable to normalise residuals and now want to be sure hw to measure model
> fit and how to report regression coefficients.
>
> Any advice out there?
>
--
Jeremy Miles
Learning statistics blog: www.jeremymiles.co.uk/learningstats
Psychology Research Methods Wiki: www.researchmethodsinpsychology.com
|