Kathy,
It's not for me to say whether the results are acceptable. You know the
research in the area. But, you suspect problems. So, is it that the odds
ratios seem large compared to the bivariate results? Is it possible that you
have collinearity issues among some of those variables? Maybe you also have
missing data and the collinearity is increased as a result of the listwise
missing data treatment. How about interactions? Perhaps the crosstabulation
of the cases actually in the analysis yields cells with none or one case.
Does adding a specific term make the CIs explode?
There's a systematic way to work this problem by beginning with the IVs that
seem to have the problem, checking results with them out of the equation to
confirm your judgement and then burrowing down into those variables to see
they relate to other IVs and to the DV. The best scecario is to be able to
identify a specific problem IV; the worst is to identify a small group.
Gene Maguin
>>I am doing multinomial logistic regression with SPSS. The problem that I
met is the large odds ratio and huge confidence interval (OR=38.215 CI:
1.736 to 843.011) for some of the variables. Other than that, everything
works fine. Is the result acceptable? If this is not acceptable, is there
any way I can fix the problem?
Thanks in advance.
Kathy
=====================
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
Locked
