Date:         Wed, 22 Jun 2005 16:58:21 +0200
Reply-To:     "nanie.payet" <nanie.payet@LAPOSTE.NET>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "nanie.payet" <nanie.payet@LAPOSTE.NET>
Subject:      Re: proc MIXED and non normal data
Comments: To: flom <flom@NDRI.ORG>
Content-Type: text/plain; charset=iso-8859-1

I want to explain a score, which represents cumulated medical costs per patient (right skewed), with the treatment recieved and  the length of stay in intensive care units.
I used proc mixed because I wanted to take into account the fact that patients are clustered into intensive care units.

thanks for your help


> >>> Payet Stéphanie 6/22/2005 10:18:55 AM >>> wrote
> <<<
> Hello all !
>
> I have fitted a random effect model with proc mixed, but the dependant
> variable isn't normaly distributed. Can I trust the results ? Are there

> other solutions ?
> For example, for the model fitted on the non normal variable, the
> treatment fixed effect isn't significant. But, if I fitt the model on the
> log-transformed variable (which can be considered as normal) the treatment has now an effect. Is it legitimate to use log-transformed data to solve the problem of normality ?
> >>>
>
> Please write back to the list (not just to me) and provide some context.....solving statistical problems without knowing the context is like boxing while blindfolded - you might hit the target, or you might break your hand.
>
> What is your DV? What are your IVs? Why would you log-transform the DV? Did you do any diagnostics? Why PROC MIXED (as opposed to GLM or whatever)
>
> In general, you could see the help for PROC MIXED, where it says that PROC MIXED assumes that the data are normally distributed;
> and, in general again, when one needs the functionality of PROC MIXED but has nonnormality, one wants to use PROC NLMIXED.
>
> HTH
>
>
> Peter
>
> Peter L. Flom, PhD
> Assistant Director, Statistics and Data Analysis Core
> Center for Drug Use and HIV Research
> National Development and Research Institutes
> 71 W. 23rd St
> www.peterflom.com
> New York, NY 10010
> (212) 845-4485 (voice)
> (917) 438-0894 (fax)

Accédez au courrier électronique de La Poste : www.laposte.net ; 
3615 LAPOSTENET (0,34€/mn) ; tél : 08 92 68 13 50 (0,34€/mn)