logistic regression

[JR]

Hello everybody,

I computed logistic regression analysis. I have four predictor variables. After including all predictors there were no significance. Which model does one prefer?

Best wishes,
JR

[Art Kendall]

Of course a lot depends on the nature of your predictors? Did you
include interactions? Did you use correct syntax, etc.?

But the short answer is:
If you you correctly did the analysis, and
if all effects fail to reach statistical significance,
then the model/practice/theory/hypothesis/policy that you prefer is the
null/default/a priori/presumed one.
Alternative model is "not proven".

Art Kendall
On 11/29/2010 7:48 AM, JR wrote:

> Hello everybody,
>
> I computed logistic regression analysis. I have four predictor variables.
> After including all predictors there were no significance. Which model does
> one prefer?
>
> Best wishes,
> JR
>
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/logistic-regression-tp3284334p3284334.html
> Sent from the SPSSX Discussion mailing list archive at Nabble.com.
>
> =====================
> 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

[parisec]

When you analyzed the bivarate association each of the predictors with the dependent variable, were any of the associations signficant?

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of JR
Sent: Monday, November 29, 2010 4:48 AM
To: [hidden email]
Subject: logistic regression

Hello everybody,

I computed logistic regression analysis. I have four predictor variables.
After including all predictors there were no significance. Which model does one prefer?

Best wishes,
JR


--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/logistic-regression-tp3284334p3284334.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

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

[JR]

Hello,

I did not include interactions. Would it be better to include them? The syntax is correct, I think. The predictors correlate with each other. Could this be the problem?  I have read that it is only a problem, if the correlation is about .80?

In the models, including only one or two predictors (each significant), the Hosmer-Lemshow Test is also significant, indicating a bad model fit? I dont know, what to do now. Would a stepwise logistic regression be better? Should I exclude predictors?

Hope, you can help me!

Best wishes,
JR

[Bruce Weaver]

JR wrote

Hello,

I did not include interactions. Would it be better to include them? The syntax is correct, I think. The predictors correlate with each other. Could this be the problem?  I have read that it is only a problem, if the correlation is about .80?

In the models, including only one or two predictors (each significant), the Hosmer-Lemshow Test is also significant, indicating a bad model fit? I dont know, what to do now. Would a stepwise logistic regression be better? Should I exclude predictors?

Hope, you can help me!

Best wishes,
JR

It's still not at all clear to me what question or questions you're trying to answer with your analysis.  Is there one?  Or are you on a fishing expedition?

Stepwise selection is almost always unwise.

    http://www.stata.com/support/faqs/stat/stepwise.html

HTH.