I don't have the context you have to be able to add anything that isn't
obvious. Being "significant" on its own is no guarantee that a predictor
belongs in a model. That's why you are doing what you are.
Otherwise my only tip is to avoid stepwise like the plague, given the
reasons often rehearsed on this list.
Nick
[email protected]
Ziad El-Khatib
Thank you and sorry for the unclear question.
Meant to:
1) know what does it mean in this case
and 2) how to deal with it?
You have kindly answered on point 1.
> The underlying reason is likely to be that you are throwing far too
many
> predictors at the response.
--> i added predictors with statistically significant value after
doing Adj. OR (2x2 table).
On Mon, Mar 2, 2009 at 12:44 PM, Nick Cox <[email protected]> wrote:
> I'm not clear what kind of answer you expect.
>
> "estimability" means being estimatable, or here strictly being _not_
> estimatable. At its simplest, the signal is thus that -logistic- found
> it too difficult to estimate the parameters in the model you
specified.
>
> The underlying reason is likely to be that you are throwing far too
many
> predictors at the response.
>
> Nick
> [email protected]
>
> Ziad El-Khatib
>
> Using -stepwise, pr(.20): logistic- I ran different variables (all of
> them and dependent variables were dichotomous); got the text below
> then got the rest in the model. Any tips what does estimability mean?
>
>
> note: incomenone dropped because of estimability
> note: awaycat dropped because of estimability
> note: busycat dropped because of estimability
> note: elsetoseecat dropped because of estimability
> note: dailyroutinecat dropped because of estimability
> note: depressedcat dropped because of estimability
> note: othercat dropped because of estimability
> note: incomeself dropped because of estimability
> note: 42 obs. dropped because of estimability
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
