Logistic regression - inflated Exp (B) and presenting results

[Katina Dimoulias]

Hi,

I have a couple of questions related to Logistic Regression that I would
appreciate assistance with.
Question 1:
I have conducted Binary Logistic regression with 6 IVs that are ordinal (5
levels) and an outcome variable (2 categories).  When I include
interactions I get Exp (B) results only on 2 variables that = 43.511 and
218.439.  Are these results acceptable or are they an indication that
something is wrong?  How can I deal with this problem?
A bit of background - I had a large number of zero frequencies when I ran
the analysis with the IVs as 5 levels and categorical.  I collapsed the 5
levels to 3 levels and ran the analysis, however still had too many zero
frequencies.  According to Hosmer and Lemeshow (2000) another strategy to
deal with this problem is, if the variable is at least ordinal scale, to
treat it as continuous.  The above results occurred when I treated the
original 5 level IVs as continuous.

Question 2:
Can anyone please suggest the convention for presenting LR analysis results
in a PhD thesis?
A number of texts suggest different results to present in summary tables of
a logistic regression analysis.  APA suggests B, SE and Odds Ratio stats,
Tabachnick & Fidell suggest B, Wald test, Odds Ratio and CIs stats.

Thank you.  Your help would be greatly appreciated.

Katina

=====================
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

[Swank, Paul R]

I would worry about quasi-separation when you have ordinal sales and
include interactions. You should get a warning about that. Another way
to check is to do cross tabs and see if you have cells with no or only
one case. As for question 2, you are better off
to over-report than under on a dissertation. However, the final
determination will be from your committee. Typically what they want, you
give.

Paul R. Swank, Ph.D
Professor and Director of Research
Children's Learning Institute
University of Texas Health Science Center
Houston, TX 77038


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Katina Dimoulias
Sent: Sunday, September 21, 2008 1:47 AM
To: [hidden email]
Subject: Logistic regression - inflated Exp (B) and presenting results

Hi,

I have a couple of questions related to Logistic Regression that I would
appreciate assistance with.
Question 1:
I have conducted Binary Logistic regression with 6 IVs that are ordinal
(5
levels) and an outcome variable (2 categories).  When I include
interactions I get Exp (B) results only on 2 variables that = 43.511 and
218.439.  Are these results acceptable or are they an indication that
something is wrong?  How can I deal with this problem?
A bit of background - I had a large number of zero frequencies when I
ran
the analysis with the IVs as 5 levels and categorical.  I collapsed the
5
levels to 3 levels and ran the analysis, however still had too many zero
frequencies.  According to Hosmer and Lemeshow (2000) another strategy
to
deal with this problem is, if the variable is at least ordinal scale, to
treat it as continuous.  The above results occurred when I treated the
original 5 level IVs as continuous.

Question 2:
Can anyone please suggest the convention for presenting LR analysis
results
in a PhD thesis?
A number of texts suggest different results to present in summary tables
of
a logistic regression analysis.  APA suggests B, SE and Odds Ratio
stats,
Tabachnick & Fidell suggest B, Wald test, Odds Ratio and CIs stats.

Thank you.  Your help would be greatly appreciated.

Katina

=====================
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