Date: Thu, 1 Apr 1999 14:46:47 -0500
Reply-To: "James B. Luther, Ph.D." <jluther@POP.NLCI.COM>
Sender: "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From: "James B. Luther, Ph.D." <jluther@POP.NLCI.COM>
Subject: Re: Appropriate design: measures of factor
Content-Type: text/plain; charset="iso-8859-1"
It would be helpful to know a little more about the purpose of the research,
I think. Is the focus on memory research (i.e. accuracy), or on other
variables in the pseudo sessions and using those to (in another experiment)
predict memory functioning?
From: Lary Jones <ljones@BINGHAMTON.EDU>
To: SPSSX-L@UGA.CC.UGA.EDU <SPSSX-L@UGA.CC.UGA.EDU>
Date: Thursday, April 01, 1999 2:34 PM
Subject: Appropriate design: measures of factor
>I really did not get a lot of feed back on the following. Let me
>re-present this with a bit more detail. I was trying to protect the
>content of the research, but I may not have provided enough detail for you
>to respond appropriately.
>A PhD student in psychology came to me with a question about how to
>impliment a design in GLM. This question is really about choices in the
>design, not so much a question of software options. For purposes of this
>question let's say that there are:
>Between Groups Factors:
> 5 (pseudo?) clinical sessions or trials
>During each one of the sesions/trials the subjects refers to a number of
>past events. Upon checking (with permission) some of these are verified to
>occur and some are not verified.
>Each of the subjects returns 4 times and similar classifications and counts
>occur in connection with each session. The dependent measures at each
>"trial" are the number of verified memories and the number of unverified
>memories, such as 10 verified and 7 unverified out of 17 (the number of
>events referred to in a session is toally free; it could be 1 or 5 or 17 or
>50, for that matter).
>It is unclear to me whether it is better to make this another W/S factor
>(verified/unverified), or treat these as multivariate "measures" taken at
>each trial. Making this a factor does allow for the direct testing of
>interactions with other factors. Treating this as "measures" means one
>would look at the discriminant function of the measures for interpretation.
>One could record this as the % of verified memories (and do the arcsin
>transformation), but this would lose the number of memories/events brought
>up, which i think is important.
>Which way would you go?
>Lary Jones % Statistical Computing Analyst
>Computing Services % ..........................
>Binghamton University % LJones@Binghamton.EDU
>Binghamton, NY 13902-6000 % (607) 777-2879