[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

From |
"Dimitriy V. Masterov" <dvmaster@lily.src.uchicago.edu> |

To |
statalist@hsphsun2.harvard.edu |

Subject |
Re: st: Re: predict using a subset of X |

Date |
Sun, 13 Jun 2004 19:33:31 -0500 (CDT) |

Thanks to Michael Blasnik for speedy help. The reason I wanted to do this is that I want to construct a residualized test score yhat that is stripped from the effect of some covariates like age. I do agree that my notation was a bit awkward. I modified his suggested code to work without manual counting: local subset "x*"; unab subset : `subset' scalar LLL=wordcount("`subset'") reg y `subset' z1 c1 d1 f1 age mat b=e(b) mat b=b[1,1..LLL] mat score yhat=b if e(sample) On Sun, 13 Jun 2004, Michael Blasnik wrote: > It's not clear what you mean here...you say you want to generate > E[y|x1,...,x50], but that would be the result from -predict- after > running -reg y x1...x50-, not running the model you show. What would you > want predict to do about the fact that there are 50 other terms in the model > that you don't want in the prediction? There is generally no reason to > think the predictions will have much to do with E[y]. > > If you just want an Xb prediction with only a subset of b, you could copy > e(b) to a matrix, create a submatrix with only the terms you want (do you > want the constant?), and then use matrix score to have Stata calculate the > "prediction". > > reg y x1...x100 > mat b=e(b) > mat b=b[1,1..50] > mat score mypredict=b > > mypredict will contain the prediction from the first 50 terms, but won't > include the remain 50 xs or the constant (which Stata puts last). > > Michael Blasnik > michael.blasnik@verizon.net > > > ----- Original Message ----- > From: "Dimitriy V. Masterov" <dvmaster@lily.src.uchicago.edu> > To: <statalist@hsphsun2.harvard.edu> > Sent: Sunday, June 13, 2004 6:57 PM > Subject: st: predict using a subset of X > > > > Is it possible to automatically generate a yhat for a subset of the RHS > > variables? > > > > Let's say I run a regression: > > > > reg y x1 x2 .... x100 > > > > I want to generate > > yhat=E[y | x1,...,x50] > > rather than > > yhat=E[y|x1,...,x100] > > > > Is there some option like predict yhat, for(x1-x50)? > > > > DVM > > > > > ____________________________________________________________________________ > ___ > > Dimitriy V. Masterov > > > > Work: > > Center for Social Program Evaluation > > 1155 East 60th St. Room 038 > > Chicago, IL 60637 > > Work: (773)256-6005 > > Fax: (773)256-6313 > > > > Home: > > 1312 East 53rd St., Apt.309 > > Chicago, IL 60615 > > Mobile: (773)220-2760 > > > * > * 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/ > _______________________________________________________________________________ Dimitriy V. Masterov Work: Center for Social Program Evaluation 1155 East 60th St. Room 038 Chicago, IL 60637 Work: (773)256-6005 Fax: (773)256-6313 Home: 1312 East 53rd St., Apt.309 Chicago, IL 60615 Mobile: (773)220-2760 * * 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/

**Follow-Ups**:**st: Heckman and collinearity: Coefficient in selection equation but nothing else?***From:*CM <cmhsieh54@yahoo.com>

**References**:**st: Re: predict using a subset of X***From:*"Michael Blasnik" <michael.blasnik@verizon.net>

- Prev by Date:
**st: Re: predict using a subset of X** - Next by Date:
**st: Heckman and collinearity: Coefficient in selection equation but nothing else?** - Previous by thread:
**st: Re: predict using a subset of X** - Next by thread:
**st: Heckman and collinearity: Coefficient in selection equation but nothing else?** - Index(es):

© Copyright 1996–2014 StataCorp LP | Terms of use | Privacy | Contact us | What's new | Site index |