Logistic Regression Interpretation Questions

[Marc Mackert]

I am working on a study that involves a one-group posttest only design.
I am using logistic regression to test a Dichotomous Dependent Variable.
 I have several Independent Variables. Specifically the study includes
seven independent variables. These IVS are all ratio, scaled from 0-1.
The analysis also includes control/intervening variables that are a
mixture of scale and categorical variables. Specifically, 1 scale and 12
categorical variables.  
 
I have entered my variables into the logistic regression model in 4
blocks. This is my first experience using logistic regression  and I am
confused by some of the results I am getting and the correct way to
interpret the results. I have several questions:
 
The first question is how high is too high for a -2LL. My -2LL is near
3000.
 
The next question is related to the interpretation of the tests of
model fit. The Omnibus Tests of  Model Coefficients are all significant
at the <p.000 level for each block. However,  the Hosmer Lemeshow test
shows model fit for blocks 1 and 4 – not for Block 2 or 3. I am confused
as to whether I should analyze the regression models 1 and 4? I will say
  my hypotheses testing is more concerned with blocks 3 and 4 than with
blocks 1 and 2.
 
Simply put, when analyzing logistic regression where variables have
been entered in as blocks, if the data do not fit the model in early
blocks – is it correct that the analyses stops there?
 
I am looking for both a reference if anyone can suggest one or
technical expertise/advice if it is available.
 
 
 
 
Marc J. Mackert, Ph.D.
Manager
Administrative Review Division
4045 S. Lowell Blvd.
Denver, Colorado 80236
Office: [phone]
Cell: [phone]
www.cdhs.state.co.us/ard

This e-mail and any attachments thereto, is intended only for use by
the addressees named herein and may contain legally privileged and/or
confidential information.  If you are not the intended recipient of this
e-mail, you are hereby notified that any dissemination, distribution or
copying of this e-mail and any attachments thereto, is strictly
prohibited.  If you have received this e-mail in error, please
immediately notify me at 303.866.7137 and permanently delete the
original and any copy of any e-mail and any printout thereof.

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

[Reutter, Alex]

Marc,

The value of the -2LL is dependent upon the size of your dataset, so the concept of "too high" doesn't apply in the absolute sense; it's mostly useful as a basis for comparing two competing models on the same set of data.

I don't think there's a quick answer for your Homser-Lemeshow Test question.  Here are references from the help system (Help>Case Studies; the recommended readings under the Logistic Regression chapter):

Hosmer, D. W., and S. Lemeshow. 2000. Applied Logistic Regression, 2nd ed. New York: John Wiley and Sons.

Kleinbaum, D. G. 1994. Logistic Regression: A Self-Learning Text. New York: Springer-Verlag.

Jennings, D. E. 1986. Outliers and residual distributions in logistic regression. Journal of the American Statistical Association, 81, 987-990.

Norusis, M. 2004. SPSS 13.0 Statistical Procedures Companion. Upper Saddle-River, N.J.: Prentice Hall, Inc.

Cheers,
Alex


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Marc Mackert
Sent: Thursday, January 24, 2008 10:54 AM
To: [hidden email]
Subject: Logistic Regression Interpretation Questions

I am working on a study that involves a one-group posttest only design.
I am using logistic regression to test a Dichotomous Dependent Variable.
 I have several Independent Variables. Specifically the study includes
seven independent variables. These IVS are all ratio, scaled from 0-1.
The analysis also includes control/intervening variables that are a
mixture of scale and categorical variables. Specifically, 1 scale and 12
categorical variables.

I have entered my variables into the logistic regression model in 4
blocks. This is my first experience using logistic regression  and I am
confused by some of the results I am getting and the correct way to
interpret the results. I have several questions:

The first question is how high is too high for a -2LL. My -2LL is near
3000.

The next question is related to the interpretation of the tests of
model fit. The Omnibus Tests of  Model Coefficients are all significant
at the <p.000 level for each block. However,  the Hosmer Lemeshow test
shows model fit for blocks 1 and 4 - not for Block 2 or 3. I am confused
as to whether I should analyze the regression models 1 and 4? I will say
  my hypotheses testing is more concerned with blocks 3 and 4 than with
blocks 1 and 2.

Simply put, when analyzing logistic regression where variables have
been entered in as blocks, if the data do not fit the model in early
blocks - is it correct that the analyses stops there?

I am looking for both a reference if anyone can suggest one or
technical expertise/advice if it is available.




Marc J. Mackert, Ph.D.
Manager
Administrative Review Division
4045 S. Lowell Blvd.
Denver, Colorado 80236
Office: [phone]
Cell: [phone]
www.cdhs.state.co.us/ard

This e-mail and any attachments thereto, is intended only for use by
the addressees named herein and may contain legally privileged and/or
confidential information.  If you are not the intended recipient of this
e-mail, you are hereby notified that any dissemination, distribution or
copying of this e-mail and any attachments thereto, is strictly
prohibited.  If you have received this e-mail in error, please
immediately notify me at 303.866.7137 and permanently delete the
original and any copy of any e-mail and any printout thereof.

=======
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