Hi Jhile
Thanks for the cautionary advice.
When I run Poisson and then run the model again with
significant predictors, I get similar results. All the
variables are significant.
However, when I run nbreg, and then run the nbreg
model again with significant predictors, I dont get
the same results,some of the variables are not
significant anymore. So I am not sure how to use the
predict command to include only the significant
predictors.
Also, when I run the predict command with nooffset
option after nbreg, the estimated values are between 0
and 1. The dependent variable in the model is
absenteeism days and I adjusted for the exposure
period using exposure () option.
My question is why are the values between 0 and 1?
What does it mean?
Am I missing something trivial here? Any thoughts?
Regards
Anju
--- [email protected] wrote:
> If you re-run the poisson model using only
> predictors that appeared
> significant in the larger model, you may find that
> the previously significant
> predictors are now not all significant. You must
> check. Simply rerun the model using
> only those predictors that were significant and
> check to see if that still
> remain the case. If it does, then type -predict mu,
> mu- to get the predicted
> counts, if that is what you want. However, I myself
> double check predictors,
> especially ones that have p-values close to .05,
> with the likelihood ratio
> test. The latter test is more reliable than Z-, t-,
> or Wald tests.
>
> Joe Hilbe
>
=========================================================
> Hi All
>
> I have a question on predict command.
>
> If I run a poisson regression such as
>
> poisson absdays male black hispanic
> bluecollarworkers
> nohealthinsurance backpain comorbidconditions
>
> where all the independent variables are dummy
> variables.
>
> absdays=constant + b1*male + b2*black + b3*hispanic
> +
> b4*bluecollarworkers + b5*nohealthinsurance +
> b6*backpain + b7*comorbidconditions
>
> If from the output only black, backpain, and
> comorbidconditions were statistically significant,
> can
> I run the predict command to obtain
> the predicted number of absdays using only the
> significant variables?
>
> In otherwords can I get the predicted absdays as
>
> constant + b2*black + b6*backpain +
> b7*comorbidconditions
>
> I know one option is to use the generate command
> but I
> would like to know if we can do this using predict
> command too.
>
> Thank you for your time.
>
> Regards
> Anju
>
>
____________________________________________________________
>
> Anju Parthan,
> Doctoral Candidate,
> The University of Texas at Austin, PHAR-PHARMACY
> ADMIN
> 1 UNIVERSITY STATION A 1930, AUSTIN, TX-78712
> Tel: (512) 459-4942; Fax: (512) 471-8762
>
>
>
>
>
>
> __________________________________
> Celebrate Yahoo!'s 10th Birthday!
> Yahoo! Netrospective: 100 Moments of the Web
> http://birthday.yahoo.com/netrospective/
> *
> * For searches and help try:
> *
> http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
> *
> * For searches and help try:
> *
> http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
____________________________________________________________
Anju Parthan,
Doctoral Candidate,
The University of Texas at Austin, PHAR-PHARMACY ADMIN
1 UNIVERSITY STATION A 1930, AUSTIN, TX-78712
Tel: (512) 459-4942; Fax: (512) 471-8762
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
