Date: Sat, 8 Apr 2006 09:53:41 -0400
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: Art Kendall <Art@DrKendall.org>
Organization: Social Research Consultants
Subject: Re: Logistic Regression
Content-type: text/plain; charset=UTF-8; format=flowed
1 (text ) * 2 (picture or not) * 2 (sound or not) * 2 (ELL or not)
Did you do a power analysis for a logistic regression looking for an
effect after controlling for several things?
What did you set as desired effect size to detect?
Do you have the large Ns necessary to do a logistic regression and
detect an effect that size?
If not do you have the ability to run more subjects?
Are there empty cells for some items in a 4 way crosstab?
The improvement in fit would be the effect that
Did it pan out in total scores in an ANOVA?
Did you first screen the text items with item analysis on a norming
sample for things like such as internal consitency, corrected
item-total, and alpha-if-item-deleted?
Are you trying to see if additional modes of presentation reduce
If so you could ENTER an average or total score on a step, then ENTER
the treatment design variables on the next step, and then the
variables that represent ELL and its interactions?
If you do not have enough cases to do a full logistic regression , you
might have to collapse the treatment design into "no supplement" vs
"supplement" for a run.
(Sound, no sound) on a run, and (picture, no picture) on a run.
Hope this helps.
Social Research Consultants
Matthew Pirritano wrote:
> These are item analyses so the DV is whether they got the item right or
> wrong (0,1). Each treatment was given to different subjects. I think
> that we have a 2 X 2 X 2 factorial design (Sound by Picture by ELL
> status) with different subjects in each cell.
> Matthew Pirritano, Ph.D.
> National Science Foundation Post-Doctoral Fellow
> College of Education
> Department of Individual, Family & Community Education
> MSC05 3040
> 1 University of New Mexico
> Albuquerque, NM 87131-0001
> Telephone (505)277-7115
> FAX (505)277-8361
> Art Kendall wrote:
>> What is/are your dependent variable(s)?
>> Do you still have access to enough subjects to try a full factorial
>> design i.e.., 2**4 design.?
>> Was each treatment given to a separate group of subjects, or were the
>> same subjects given all of the treatments.
>> Social Research Consultants
>> Matthew Pirritano wrote:
>>> Hello all,
>>> We've conducted a study where we've tested the effects of including
>>> narration and pictures with math items. The narration is simply a
>>> reading aloud the problem. There were four conditions (Text Only,
>>> Picture and Text, Picture and Narration, Sound and Text). We divided
>>> our analyses into two main conditions, Picture (with picture, without
>>> picture), and Sound (with narration, without narration). I've run
>>> logistic regressions (LR) looking to see if the effects of our
>>> conditions varied with language groups (ELL or non-ELL), while
>>> controlling for overall performance on all of the problems (11
>>> problems). I was originally told to run the logistic regressions
>>> including either the Picture variable, OR the Sound variable. I did
>>> this. But it occurred to me that by only including one of these
>>> variables at a time I was ignoring the fact that a problem could have
>>> had both Picture and Sound present or neither present. Should I have
>>> included Picture and Sound in the LR simultaneously? And because I'm
>>> interested in the interaction between condition and language group
>>> is it
>>> okay to run one set of analyses where I include the interaction between
>>> picture and language group, and another including the interaction
>>> between sound and language group? My reasoning for the two sets of
>>> analyses is that if I included both interactions at the same time the
>>> results would not be very interpretable.
>>> As always, any help much appreciated.
>>> Matthew Pirritano, Ph.D.
>>> National Science Foundation Post-Doctoral Fellow
>>> College of Education
>>> Department of Individual, Family & Community Education
>>> MSC05 3040
>>> 1 University of New Mexico
>>> Albuquerque, NM 87131-0001
>>> Telephone (505)277-7115
>>> FAX (505)277-8361