Hi Brian,

1) You obtain the best possible transformation after

RANK VARIABLES= myvar /NORMAL .

(Write your own variable name instead of myvar.) You can see the resulting transformation after

GRAPH /SCATTERPLOT(BIVAR)= myvar WITH Nmyvar .

The problem here is that you get a function without an explicit formula and therefore not able to generalize directly - an approximate formula can be estimated using e.g. non-linear regression, but this can be a cumbersome way, I think.

2) I routinely use a macro which I have once published here. It is a heuristic algorithm checking many possible transformations and choosing the "best" one. If you wish, I will send it off-list to you. Of course no guarantee of something, but for my data mining purposes it works better than the standard transformation-seeker in the SAS data mining suit which I currently use.

3) Perhaps you know it, but I should nevertheless add a warning: Be very cautious. You can often find something similar to the spurious correlations which has been so nicely demonstrated by Marta García-Granero in her recent postings. One or two outliers or erroneous cases can change your formula very much. And it is always dangerous to extrapolate the non-linear transformation functions behind the scope over which they were developed.

Greetings

Jan

-----Original Message----- From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU]On Behalf Of Brian McMillan Sent: Monday, June 13, 2005 2:13 PM To: SPSSX-L@LISTSERV.UGA.EDU Subject: Re: non-normal distributions

Hello all,

Maybe I'm searching for the Holy Grail here but does anyone know if there is any software available for SPSS which will calculate the best transformation to use in order to transform your data closer to a normal distribution?

Thanks in advance, Brian