Nor did I recommend that.
predict glm, res
is what I recommended knowing that -res-
is an abbreviation for -response-. You
should be able to work that kind of
detail out for yourself by reading the help.
Nick
[email protected]
Cinzia Rienzo
> Thanks nick,
> but with glm STATA does not allow : predict glm, resid!
> I have to check
> Nick Cox wrote:
>
>
> >No, it is not correct. You want
> >
> >predict resglm, res
> >
> >as Maarten implied.
> >
> >Nick
> >[email protected]
> >
> >Cinzia Rienzo
> >
> >> thanks Marteen,
> >> for glm i used: predict resglm
> >> for ols i use: predict resols, resid
> >>
> >> is this correct?
> >> but when i use the predict of Glm i get (almost) the same result!
> >
> >Maarten buis wrote:
> >
> >> >Looking at the resglm results I don't think they are the
> >> residuals but
> >> >the linear predictor. Did you correctly use -predict- here
> >> (the options
> >> >of predict after -glm- are different than those after -regress-).
> >
> >Cinzia Rienzo
> >
> >> >> i am replicating a paper so i know the results i should
> >> get; the aim
> >> >> is
> >> >> to analyse the residual of a regression;
> >> >> the problem is that when i regress the following:
> >> >>
> >> >> glm lwage1 age educ educage educage2 educage3
> educage4 [aw=hwt]
> >> >>
> >> >> i obtain the same results as the author exept for the
> >> variance gap of
> >> >>
> >> >> the 90-10 centile.
> >> >>
> >> >> When instead i regress the following:
> >> >> reg lwage1 age educ educage educage2 educage3
> educage4 [aw=hwt]
> >> >> i get the same 90-10 gap centile but the other results are
> >> different.
> >> >>
> >> >> Both regressions give the same coefficient but the residual are
> >> >> different:
> >> >>
> >> >> Variable | Obs Mean Std. Dev. Min
> Max
> >> >>
> >>
> ------------+--------------------------------------------------------
> >> >> resols | 1761676 -.0012382 .454438 -2.421267
> 3.340183
> >> >> resglm | 1761676 1.77811 .3409537 .6651884 2.437594
> >> >>
> >> >> Could you please suggest me what could be the problem or is it
> >> >> methodologically not correct to work with the two
> regressions and
> >> >> therefore with the two residuals?
*
* 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/
