SPSSX Discussion

Guide to Binary Logistic Regression

Classic

List

Threaded

5 messages Options

Cardiff Tyke

Guide to Binary Logistic Regression

Hi,

Can anyone point me to a guide to how to prepare datasets for Binary Logistic Regression. Currently I just take the scattergun approach of adding my dependent variable and running it against every single covariate I have to hand. The upshot is that I always seem to get poor results from my observed results and predicted probabilities despite the sample file containing almost a 50:50 split of good to bad.

I appreciate that this can indicate my independent variables are of little use but I can't help feeling that I should be doing more with my analysis to confirm that this is the case.

Regards,
JC

___________________________________________________________
Yahoo! Answers - Got a question? Someone out there knows the answer. Try it
now.
http://uk.answers.yahoo.com/

Muir Houston

Re: Guide to Binary Logistic Regression

Not knowing what you are investigating i.e. the DV and the explanators it is difficult to comment.
What does existing research in the area of interest suggest are the important explanators?
Are you using for example the change in the chi-square as measure to decide whether or not to include variables (preferably guided by theory)

Muir Houston
Research Fellow
CRLL
Institute of Education
University of Stirling
FK9 4LA
01786-46-7615

________________________________

From: SPSSX(r) Discussion on behalf of Cardiff Tyke
Sent: Wed 04/07/2007 11:31
To: [hidden email]
Subject: Guide to Binary Logistic Regression

Hi,

Can anyone point me to a guide to how to prepare datasets for Binary Logistic Regression. Currently I just take the scattergun approach of adding my dependent variable and running it against every single covariate I have to hand. The upshot is that I always seem to get poor results from my observed results and predicted probabilities despite the sample file containing almost a 50:50 split of good to bad.

I appreciate that this can indicate my independent variables are of little use but I can't help feeling that I should be doing more with my analysis to confirm that this is the case.

Regards,
JC

___________________________________________________________
Yahoo! Answers - Got a question? Someone out there knows the answer. Try it
now.
http://uk.answers.yahoo.com/

--
The University of Stirling is a university established in Scotland by
charter at Stirling, FK9 4LA. Privileged/Confidential Information may
be contained in this message. If you are not the addressee indicated
in this message (or responsible for delivery of the message to such
person), you may not disclose, copy or deliver this message to anyone
and any action taken or omitted to be taken in reliance on it, is
prohibited and may be unlawful. In such case, you should destroy this
message and kindly notify the sender by reply email. Please advise
immediately if you or your employer do not consent to Internet email
for messages of this kind.

Cardiff Tyke

Re: Guide to Binary Logistic Regression

In reply to this post by Cardiff Tyke

For a marketing exercise, I am trying to predict homeownership and I have around 50 demographic variables at my disposal and a sample dataset of around 13,000 homeowners/non homeowners.

The only saving grace of my poor results is that an list-members based in the Midlands (UK) will get 1 less piece of junk mail in a few months time!

----- Original Message ----
From: Muir Houston <[hidden email]>
To: [hidden email]
Sent: Wednesday, 4 July, 2007 11:43:23 AM
Subject: Re: Guide to Binary Logistic Regression

Not knowing what you are investigating i.e. the DV and the explanators it is difficult to comment.
What does existing research in the area of interest suggest are the important explanators?
Are you using for example the change in the chi-square as measure to decide whether or not to include variables (preferably guided by theory)

Muir Houston
Research Fellow
CRLL
Institute of Education
University of Stirling
FK9 4LA
01786-46-7615

________________________________

From: SPSSX(r) Discussion on behalf of Cardiff Tyke
Sent: Wed 04/07/2007 11:31
To: [hidden email]
Subject: Guide to Binary Logistic Regression

Hi,

Can anyone point me to a guide to how to prepare datasets for Binary Logistic Regression. Currently I just take the scattergun approach of adding my dependent variable and running it against every single covariate I have to hand. The upshot is that I always seem to get poor results from my observed results and predicted probabilities despite the sample file containing almost a 50:50 split of good to bad.

I appreciate that this can indicate my independent variables are of little use but I can't help feeling that I should be doing more with my analysis to confirm that this is the case.

Regards,
JC

___________________________________________________________
Yahoo! Answers - Got a question? Someone out there knows the answer. Try it
now.
http://uk.answers.yahoo.com/

--
The University of Stirling is a university established in Scotland by
charter at Stirling, FK9 4LA. Privileged/Confidential Information may
be contained in this message. If you are not the addressee indicated
in this message (or responsible for delivery of the message to such
person), you may not disclose, copy or deliver this message to anyone
and any action taken or omitted to be taken in reliance on it, is
prohibited and may be unlawful. In such case, you should destroy this
message and kindly notify the sender by reply email. Please advise
immediately if you or your employer do not consent to Internet email
for messages of this kind.

___________________________________________________________
New Yahoo! Mail is the ultimate force in competitive emailing. Find out more at the Yahoo! Mail Championships. Plus: play games and win prizes.
http://uk.rd.yahoo.com/evt=44106/*http://mail.yahoo.net/uk

ViAnn Beadle

Re: Guide to Binary Logistic Regression

This sounds like a classic cart application implemented in SPSS as the TREE
procedure.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Cardiff Tyke
Sent: Wednesday, July 04, 2007 4:53 AM
To: [hidden email]
Subject: Re: Guide to Binary Logistic Regression

For a marketing exercise, I am trying to predict homeownership and I have
around 50 demographic variables at my disposal and a sample dataset of
around 13,000 homeowners/non homeowners.

The only saving grace of my poor results is that an list-members based in
the Midlands (UK) will get 1 less piece of junk mail in a few months time!

----- Original Message ----
From: Muir Houston <[hidden email]>
To: [hidden email]
Sent: Wednesday, 4 July, 2007 11:43:23 AM
Subject: Re: Guide to Binary Logistic Regression

Not knowing what you are investigating i.e. the DV and the explanators it is
difficult to comment.
What does existing research in the area of interest suggest are the
important explanators?
Are you using for example the change in the chi-square as measure to decide
whether or not to include variables (preferably guided by theory)

Muir Houston
Research Fellow
CRLL
Institute of Education
University of Stirling
FK9 4LA
01786-46-7615

________________________________

From: SPSSX(r) Discussion on behalf of Cardiff Tyke
Sent: Wed 04/07/2007 11:31
To: [hidden email]
Subject: Guide to Binary Logistic Regression

Hi,

Can anyone point me to a guide to how to prepare datasets for Binary
Logistic Regression. Currently I just take the scattergun approach of
adding my dependent variable and running it against every single covariate I
have to hand. The upshot is that I always seem to get poor results from my
observed results and predicted probabilities despite the sample file
containing almost a 50:50 split of good to bad.

I appreciate that this can indicate my independent variables are of little
use but I can't help feeling that I should be doing more with my analysis to
confirm that this is the case.

Regards,
JC

___________________________________________________________
Yahoo! Answers - Got a question? Someone out there knows the answer. Try it
now.
http://uk.answers.yahoo.com/

--
The University of Stirling is a university established in Scotland by
charter at Stirling, FK9 4LA. Privileged/Confidential Information may
be contained in this message. If you are not the addressee indicated
in this message (or responsible for delivery of the message to such
person), you may not disclose, copy or deliver this message to anyone
and any action taken or omitted to be taken in reliance on it, is
prohibited and may be unlawful. In such case, you should destroy this
message and kindly notify the sender by reply email. Please advise
immediately if you or your employer do not consent to Internet email
for messages of this kind.

___________________________________________________________
New Yahoo! Mail is the ultimate force in competitive emailing. Find out more
at the Yahoo! Mail Championships. Plus: play games and win prizes.
http://uk.rd.yahoo.com/evt=44106/*http://mail.yahoo.net/uk

Hector Maletta

Re: Guide to Binary Logistic Regression

In reply to this post by Cardiff Tyke

Getting poor results from you're a priori predictors is a common
occurrence, and a valid scientific result in itself (there is indeed a
regrettable so-called 'publication bias' whereby researchers publish only
positive results, thus masking the true picture).
Another warning is that using Log Reg to predict the outcome of
individual cases, and using this as a criterion (as in the 'classification
table' produced by SPSS, crossing actual and predicted outcome) is not
generally right. Log Reg does not predict outcomes, but the probability of
outcomes. Probabilities are not observable at the level of individuals.
Moreover, an individual can produce the outcome even having a low
probability, and the converse. Probabilities are predicated of populations,
not of individuals, so you would observe that the outcome is relatively more
frequent in populations or groups with higher predicted probability
(although perhaps not happening exactly to those individuals with the
highest individual probability).
This may help you discover that your results are not so
discouraging after all. For more detailed and substantive advice you should
provide more detail. Anyway, the choice of predictors is a matter of theory
and specific expertise, not of statistical wisdom, though SPSS provides the
Stepwise method to try out a large set of would-be predictors and select the
most effective. Various objections and qualifications have been moved
against Stepwise, but you may want to consider it.

Hector

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Cardiff Tyke
Sent: 04 July 2007 07:31
To: [hidden email]
Subject: Guide to Binary Logistic Regression

Hi,

Can anyone point me to a guide to how to prepare datasets for
Binary Logistic Regression. Currently I just take the scattergun approach
of adding my dependent variable and running it against every single
covariate I have to hand. The upshot is that I always seem to get poor
results from my observed results and predicted probabilities despite the
sample file containing almost a 50:50 split of good to bad.

I appreciate that this can indicate my independent variables are of
little use but I can't help feeling that I should be doing more with my
analysis to confirm that this is the case.

Regards,
JC

___________________________________________________________
Yahoo! Answers - Got a question? Someone out there knows the
answer. Try it
now.
http://uk.answers.yahoo.com/