Date:         Mon, 17 Jun 1996 19:01:13 GMT
Reply-To:     David Nichols <nichols@SPSS.COM>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         David Nichols <nichols@SPSS.COM>
Organization: SPSS, Inc.
Subject:      Re: Tired of waiting . . .

In article <31C1B747.2DDD@psych.uni-halle.de>,
Sven Blankenberger  <s.blankenberger@psych.uni-halle.de> wrote:
>Jerry Dallal wrote:
>
>>             Sometimes repeated measures are not longitudinal and
>>         sometimes the sphericity condition is met and sometimes both
>>         things happen simultaneously, or is am MCP invalid even under
>>         these conditions?  I hear rumors that 2-nd tier packages are
>>         starting to include these tests.  Why not the big guys?  I'd
>>         prefer not having to add another package to my machine if
>>         possible, but I'm getting tired of writing DATA steps to
>>         rearrange my data.
>
>In some way I'm happy to read this post: I'm not the only one who wants
>MCPs for within designs. I've almost always complete within designs
>(three or four within factors, no between factor) and it is troublesome
>not to have a simple way to do post-hoc analyses (pairwise comparisons
>or complex contrasts). The interface for post-hoc analysis in between designs
>is quite well and I would like to have something similar for within designs
>(with an option for seperate vs. pooled error terms).
>
>Another issue in within designs: graphing the data (e.g. line plot of means)
>requires always a complete rearrangement of the data files. For an anova
>the responses from each subject compromise one (long) row (exception: using
>mglh and specifying each effect and error term and so on; wast of time
>for standard analyses). But for graphics I need a data layout as if it
>is a between design. If I remember correctly, V6 includes a REPEAT option,
>but this works only for a single within factor. It would be helpful to have
>a REPEAT option for graphics similar to the repeat option for an anova, but
>with more flexibility. Most of the time it is not obvious (at least for me)
>which variable (within factor) should be placed on the x-axis and which one
>should be represented by different lines. Thus, it is good practice to play
>around for finding a good solution. At the moment this requires between-layout
>of the data.
>
>And, as stated half a year ago: please think about including another pair of
>ymin, ymax options with the version 5 behavior, i.e. restricting the range of
>the y-axis but NOT excluding data. The version 6 approach seems appropriate
>for exploratory data analysis but not for a final graph (at least for me).
>Ok, I could save the means and use this file for the plot I want, but: I like
>to have as few data files as possible for a given experiment.
>I know, Leland, we disagree on this point. But maybe it is possible to give
the
>users the choice.
 
I already discussed some of the issues that complicate this with within
subjects design in a previous posting, but did note that automatic
pairwise comparisons without special error adjustments, using separate
error terms for within subjects comparisons, are coming to SPSS GLM in
release 7.5. Profile or interaction plots for between and/or within
factors are already available point and click in GLM in release 7.0.
 
--
-----------------------------------------------------------------------------
David Nichols             Senior Support Statistician              SPSS, Inc.
Phone: (312) 329-3684     Internet:  nichols@spss.com     Fax: (312) 329-3668
-----------------------------------------------------------------------------