Discriminant Function Analysis

[Kathryn Gardner]

Dear List,
 
I have a few questions about multiple discriminant function analysis (DFA) that I hope people can help me with.
 
1) I have seen people report the results of a MANOVA (as a preliminary exploration of group differences) prior to running a DFA. Is this standard?
2) I have three groups in my dependent variable (Borderline Personality Disorder [BPD]) and my predictors are a set of emotional intelligence (EI) variables. BPD and EI are negatively correlated when running a Pearsons correlation. Similarly, in the DFA SPSS output, the mean EI scores are much lower in the BPD as opposed to possible BPD and non-BPD group. However, in the output, the "Standardized Canonical Discriminant Function Coefficients" box shows that my EI variables have positive signs. I don't follow this, as I assume this means that as my Emotional Intelligence variables increase so does BPD which isn't the case. Can anyone shed any light on this?
3) I want to perform hierarchical DFA and have read that the ANALYSIS syntax command can be used.  To run a hierarchical DFA, is it correct to first run the analysis excluding the control variables (in my case age and gender) then run it with all variables in? The results of the two can then be compared? Or is there another technique? I am hoping that the syntax below does this correctly??? Thanks Kathryn.
 
DISCRIMINANT
/GROUPS=BPD_code2(0 3)
/VARIABLES=gender age optreg appraisal socskills utiliz
/ANALYSIS gender age
/ANALYSIS all
/PRIORS SIZE
/STATISTICS=MEAN STDDEV UNIVF TABLE
/CLASSIFY=NONMISSING POOLED .
 
I'd appreciate any help with this.
_________________________________________________________________
Try Live.com - your fast, personalized homepage with all the things you care about in one place.
http://www.live.com/getstarted

[Fry, Jonathan B.]

Re 2: The direction of the canonical discriminant function is selected to maximize the number of positive coefficients.  To see whether high or low scores show up in a particular group, look at the group's centroid.  That's the mean canonical discriminant score in the group.

Re 3: I'm sure the results would be much more interesting if the hierarchical analysis were done in the other order (demographics first, then variables of interest).

Jonathan Fry
SPSS Inc.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Kathryn Gardner
Sent: Monday, April 23, 2007 8:53 AM
To: [hidden email]
Subject: [BULK] Discriminant Function Analysis
Importance: Low

Dear List,

I have a few questions about multiple discriminant function analysis (DFA) that I hope people can help me with.

1) I have seen people report the results of a MANOVA (as a preliminary exploration of group differences) prior to running a DFA. Is this standard?
2) I have three groups in my dependent variable (Borderline Personality Disorder [BPD]) and my predictors are a set of emotional intelligence (EI) variables. BPD and EI are negatively correlated when running a Pearsons correlation. Similarly, in the DFA SPSS output, the mean EI scores are much lower in the BPD as opposed to possible BPD and non-BPD group. However, in the output, the "Standardized Canonical Discriminant Function Coefficients" box shows that my EI variables have positive signs. I don't follow this, as I assume this means that as my Emotional Intelligence variables increase so does BPD which isn't the case. Can anyone shed any light on this?
3) I want to perform hierarchical DFA and have read that the ANALYSIS syntax command can be used.  To run a hierarchical DFA, is it correct to first run the analysis excluding the control variables (in my case age and gender) then run it with all variables in? The results of the two can then be compared? Or is there another technique? I am hoping that the syntax below does this correctly??? Thanks Kathryn.

DISCRIMINANT
/GROUPS=BPD_code2(0 3)
/VARIABLES=gender age optreg appraisal socskills utiliz /ANALYSIS gender age /ANALYSIS all /PRIORS SIZE /STATISTICS=MEAN STDDEV UNIVF TABLE /CLASSIFY=NONMISSING POOLED .

I'd appreciate any help with this.
_________________________________________________________________
Try Live.com - your fast, personalized homepage with all the things you care about in one place.
http://www.live.com/getstarted