SPSS Syntax for Mixed Design Simple Effects

[Dina]

Hi there,
I have searched extensively for the SPSS syntax required for me to tease
apart a significant three way interaction in an A X (B) X (C) Mixed
Factorial Design, where A is between subjects and B and C are
within/repeated. All the examples I have come across explaining how to do
this with MANOVA syntax were much simpler than my case, where A has 3 levels
(Depression group) B has 9 levels (9 different people) and C has 19 levels
(19 different trait dimensions).  I'd like to examine the simple simple main
effects of A within each level of B and within each level of C and cannot
figure out how to do this in SPSS syntax (e.g, do people with clinical
depression versus some depression versus normals self-present differently to
(each of the 9 targets) on (each of the 19 traits)?)
Any help at all would be GREATLY appreciated!
Thanks so much,
Dina

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Ryan]

Dina,

The TEST subcommand offered in MIXED should be capable of estimating
the simple effects you desire. First thing's first, though...You must
properly parameterize the model. I believe I recently posted a
solution for parameterizing a similar model here:

http://www.listserv.uga.edu/cgi-bin/wa?A2=ind1104&L=spssx-l&P=R23020

If you are interested in fitting this model via the MIXED procedure
and would like some assistance after reading up on linear MIXED models
(including the post above), write back and I will try to help.

Best wishes,

Ryan

On Thu, Apr 28, 2011 at 5:44 PM, Dina <[hidden email]> wrote:

> Hi there,
> I have searched extensively for the SPSS syntax required for me to tease
> apart a significant three way interaction in an A X (B) X (C) Mixed
> Factorial Design, where A is between subjects and B and C are
> within/repeated. All the examples I have come across explaining how to do
> this with MANOVA syntax were much simpler than my case, where A has 3 levels
> (Depression group) B has 9 levels (9 different people) and C has 19 levels
> (19 different trait dimensions).  I'd like to examine the simple simple main
> effects of A within each level of B and within each level of C and cannot
> figure out how to do this in SPSS syntax (e.g, do people with clinical
> depression versus some depression versus normals self-present differently to
> (each of the 9 targets) on (each of the 19 traits)?)
> Any help at all would be GREATLY appreciated!
> Thanks so much,
> Dina
>
> =====================
> To manage your subscription to SPSSX-L, send a message to
> [hidden email] (not to SPSSX-L), with no body text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send the command
> INFO REFCARD
>

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Rich Ulrich]

As I read this, you have 9 subjects, who are each classed into
one of 3 groups; and you want to analyze -- in one analysis --
19 traits.

Do I understand this correctly?  - If that is it, I would suggest
that about the only MAIN analysis possible is to take one score,
selecting it from the 19 or computing a composite for several
(or all) of them;  and look at either the correlation (if the
groups are ordered) or the simple ANOVA.  Your power of analysis
will be tiny, for 3 groups with N=9,  but any other analysis
would be even smaller ... if it is even interpretable.

If that is not it, please clarify.

--
Rich Ulrich




=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Dina Gohar]

Hi Rich,

Thanks so much for trying to help! Sorry for the confusion. I have over 150 subjects so power should not be an issue. It is all within subjects [no experimentally manipulated conditions], though people have been classified into 1 of 3 groups based on their depression scores, which is one of the predictor variables of interest.

Each of the subjects filled out a Likert Scale for 19 trait dimensions 9 times, one for each of the 9 people or targets [e.g., their mom, their best friend, an authority figure, a child.], as if they were going to show those ratings to that person/target. One of the goals here is to look at whether people who score high/low/0 on depression self-present on any of these traits to any of these targets differently. The three way interaction between depression trait and target was significant in my mixed ANOVA, I am just having trouble figuring out how to parse it further [beyond finding out that the two-way interaction is significant within each level of depression] with simple effects in SPSS to find out the exact effects of depression within each target and within each trait.

I hope this clarification made sense. Thanks again for any suggestions!

Dina

On Fri, Apr 29, 2011 at 2:21 AM, Rich Ulrich <[hidden email]> wrote:

As I read this, you have 9 subjects, who are each classed into
one of 3 groups; and you want to analyze -- in one analysis --
19 traits.

Do I understand this correctly?  - If that is it, I would suggest
that about the only MAIN analysis possible is to take one score,
selecting it from the 19 or computing a composite for several
(or all) of them;  and look at either the correlation (if the
groups are ordered) or the simple ANOVA.  Your power of analysis
will be tiny, for 3 groups with N=9,  but any other analysis
would be even smaller ... if it is even interpretable.

If that is not it, please clarify.

--
Rich Ulrich


                                         

[Ryan]

If N=9, then clearly the approach I suggested would be inappropriate.
If the OP wants help, he/she needs to write back with a clarification
around sample size.

Ryan

On Fri, Apr 29, 2011 at 2:21 AM, Rich Ulrich <[hidden email]> wrote:

> As I read this, you have 9 subjects, who are each classed into
> one of 3 groups; and you want to analyze -- in one analysis --
> 19 traits.
>
> Do I understand this correctly?  - If that is it, I would suggest
> that about the only MAIN analysis possible is to take one score,
> selecting it from the 19 or computing a composite for several
> (or all) of them;  and look at either the correlation (if the
> groups are ordered) or the simple ANOVA.  Your power of analysis
> will be tiny, for 3 groups with N=9,  but any other analysis
> would be even smaller ... if it is even interpretable.
>
> If that is not it, please clarify.
>
> --
> Rich Ulrich
>
>
>
>
> =====================
> To manage your subscription to SPSSX-L, send a message to
> [hidden email] (not to SPSSX-L), with no body text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send the command
> INFO REFCARD
>

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD