regressions

[shahd sal]

Hi everyone.
I'm doing a logistic regression for 7 dependent variables and 3 independent values. The total data set is 1646.
I chose logistic regression since the dependent variables had 2 values (yes, no) "I excluded the (Unknown) values ".

after I did the regression I noticed many observations were omitted in one of the independent variables (age groups), as it written in stata, some because of "predicts failure perfectly" and some " omitted because of collinearity."

could you please help me understand what does these means? is it better to change to a different regression model?

And I also need to do the adjusted an unadjusted estimates of association.. would you please explain how it should be done ?

would really appreciate the help

[Rich Goldstein]

this is discussed in the manual entry for -logit- in r.pdf

[Carlo Lazzaro]

Shahd:
just out of curiosity: how can you run a -logit- (or -logistic-) with 7 dependent variables?

[shahd sal]

hi, by doing each dependent variable separately. Is this not the way to do it?

[Carlo Lazzaro]

Shahd:
thanks for clarifying.
Yes, you can run 7 different logistic regressions, but I fear you will have hard times in disseminating your results.
In addition, the scant number of (three) predictors (that, seem to be the same for each regression, too) does not seem to help either.
At the toop of that, your predictors seems to have a limited variation ("predicts failure perfectly)and suffer from perfect correlation too (basically, Stata finds impossible to tease out the contribution of each one of them to variation in the dependent variable).
That said, I do echo's Rich's wise advice to take a (comprehensive) look at -logit- entry in Stata .pdf manual and related references.