Date: Tue, 4 May 2004 16:24:54 +1000
Reply-To: paulandpen@optusnet.com.au
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: Paul Dickson <paulandpen@optusnet.com.au>
Subject: Re: log transformation and repeated measures
Content-Type: text/plain
Katherine,
I agree with Paul that you need to transform all or none of the time series stages.
As for the transformation, that is the particular challenge. If you do transform, then
the original meaning of the scale is lost (trading off-statistical appropriatness vs
scale meaning). I always err on the side of meaning myself. As for the breach, if
the largest group has sd measures that are three times the smallest group
(assuming unequal sample sizes) than I would consider this a problem and maybe
worth tranforming. The analysis you are talking about is fairly robust to this
breach, particularly when the sample sizes of the groups are the same. I am
confused as to whether you are talking about a breach of sphericity (there is an
epsilon correction for this in SPSS) given that it is repeated measures. Finally,
from what I remember, the breach in homogeneity creates a greater chance of
type 1error, so that a small correction in alpha (tightening) may be justified as a
less complex solution.
HTH
Regards Paul
> Paul R Swank <Paul.R.Swank@uth.tmc.edu> wrote:
>
> Do you violate homogeneity of variance between groups? Or sphericity?
> It
> would be incorrect to transform two of the three variables in a
> repeated
> measures design. But more information is needed on how non-normal the
> data
> is and in what way. There may be a way to handle it with a mixed
> model but
> it will depend on what your research questions are and what the data
> looks
> like.
>
> Paul R. Swank, Ph.D.
> Professor, Developmental Pediatrics
> Medical School
> UT Health Science Center at Houston
>
>
> -----Original Message-----
> From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] On Behalf
> Of
> Grimm, Katherine
> Sent: Monday, May 03, 2004 1:12 PM
> To: SPSSX-L@LISTSERV.UGA.EDU
> Subject: log transformation and repeated measures
>
>
> Good Monday afternoon,
>
> I have a case with 3 repeated measures. All three measures are
> measured on
> a scale from 1 to 10 at time 1, time 2 and time 3. The time 1 is
> normally
> distributed, but the second and third are not. When I run repeated
> measures, I violate homogeneity of variance. I believe that I should
> do a
> log transformation with the latter two, but then what about the
> first? How
> do I interpret this when I look at the repeated measures analysis?
>
> Thanks in advance for your insight.
>
> Katherine Grimm, MPH
> Statistician
> HealthEast Care System
> Research and Education - Midway
> 1700 University Ave W
> St. Paul, MN 55104-3727
>
> phone: 651-232-5261
> fax: 651-641-0683
>
>
>
> The information included in this e-mail message, including any
> attachments,
> is intended only for the person or organization to which it is
> addressed.
> This e-mail message may contain information that is privileged or
> confidential. If you receive this e-mail message and are not the
> intended
> recipient or responsible for delivering the message to the intended
> recipient, you may not use, disseminate, distribute or copy the
> information
> included in this e-mail and any attachments. If you received this
> e-mail
> message by mistake, please reply by e-mail and destroy all copies of
> this
> message and any attachments. Thank you.
| 
