Date:         Mon, 12 Dec 2011 20:44:31 +0000
Reply-To:     "Poes, Matthew Joseph" <mpoes@UILLINOIS.EDU>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         "Poes, Matthew Joseph" <mpoes@UILLINOIS.EDU>
Subject:      Re: pretest analysis
Comments: To: Rich Ulrich <rich-ulrich@live.com>
In-Reply-To:  <BLU143-W20DE924979A23C21531D4997BC0@phx.gbl>
Content-Type: multipart/alternative;

I would think that giving the same test at multiple time points to the same population would create the strong possibility of a testing effect, and that this could lead to potentially "better" scores than for those who did not have the opportunity for a pre-test.  Since this does not appear to be a pre-post design of any sort, the difference in growth would be a non-issue, instead you would just end up with a potential group of 15 who have scores which systematically differ due to prior exposure to the test/survey/stimuli.  For me, this would make an indicator variable mandatory even for fairly casual reporting and comparison, especially if there was evidence that the pre-test group was systematically different to begin with (i.e. the people who received the pre-test were a different "kind" of person, say the first that showed up).

Matthew J Poes
Research Data Specialist
Center for Prevention Research and Development
University of Illinois

From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] On Behalf Of Rich Ulrich
Sent: Monday, December 12, 2011 2:30 PM
To: SPSSX-L@LISTSERV.UGA.EDU
Subject: Re: pretest analysis

Being in the pretest group might *make* a difference, as Paul suggests -
You can check that further by looking at pre-post scores for the 15 (assuming
that the same test was given later).  A paired t-test will have more power
than a student's t-test when there is a high correlation between scores.

Or being in the pretest group might reflect some existing differences between
the subjects on demographic or other factors, which may be correlated with
who received a pre-test... .
> Date: Mon, 12 Dec 2011 10:47:52 -0600
> From: Paul.R.Swank@uth.tmc.edu<mailto:Paul.R.Swank@uth.tmc.edu>
> Subject: Re: pretest analysis
> To: SPSSX-L@LISTSERV.UGA.EDU<mailto:SPSSX-L@LISTSERV.UGA.EDU>
>
> If you categorize the sample into two groups, those with the pretest and those without, then you can compare the groups to see if having a pretest made a difference. However, with 15 subjects in the pretest group, you will likely not have enough power to adequately test that hypothesis. You can't use the pretest score itself in a model since the rest of the sample didn't have it.
>
[snip]

--
Rich Ulrich


[text/html]