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

From |
frone@ria.buffalo.edu |

To |
statalist@hsphsun2.harvard.edu |

Subject |
Re: st: offset variables |

Date |
Thu, 4 Jul 2002 19:06:30 -0400 |

Sam, Thanks much. The explanation was helpful. It gives me a place to begin working through this in more detail for myself. Mike SamL <saml@demog.berkeley.edu> Sent by: owner-statalist@hsphsun2.harvard.edu 07/04/02 07:43 AM Please respond to statalist To: statalist@hsphsun2.harvard.edu cc: SamL@demog.berkeley.edu Subject: Re: st: offset variables I cannot provide an exhaustive set of reasons for why one would use the offset feature rather than enter the exposure as a covariate. But my understanding is the following. By constraining the coefficient of the exposure variable to equal one you transform the model into a model of rates (e.g., injuries per unit of exposure instead of the probability that the person will be injured). This occurs because the link function is logistic, and one implication of that is that a positive coefficient of 1, subtracted from both sides of the equation, is equivalent to making the exposure the denominator of the left-hand side of the equation. Further, if you estimathe the coefficient (which should be unnecessary if you know that you want to transform the dependent variable into a rate), you would obtain a standard error and, as the term is estimated in the model, the variance-covariance matrix of the estimates will reflect the association between this estimate and other estimates in the model. In other words, the precision of other estimates will be impacted by estimating this coefficient. If the reason for including the variable as an exposure is just to make the model a model of rates, then there is no need to waste information by estimating the coefficient and thereby decreasing the precision of other estimates. At least, that is my understanding. Hope this helps. Sam On Mon, 1 Jul 2002 frone@ria.buffalo.edu wrote: > I'm reading a matched case-control study that used conditional logistic > regression (in STATA) to explore the predictors of being injured at work > during the prior 12 months. The analysis uses number of days worked > during the period as an offset variable presumably to adjust for > differential exposure. In conditional logistic regression the coefficient > for the offset is constrained to equal 1. > > I don't understand why one would make this constraint, as opposed to using > number of days worked as another predictor variable. Is there a reference > that explains the various uses for offset variables? I wasn't able to > find a useful explanation in the STATA archives or through a general > search of the net. > > Thanks in advance for any help. > > Mike Frone > * > * 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/ * * 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: reduced form***From:*"Stephanie Zobay" <szobay@directvinternet.com>

- Prev by Date:
**st: RE: RE: RE: RE: RE: RE: Re: RE: may not use time-series operators on string variables** - Next by Date:
**st: reduced form** - Previous by thread:
**Re: st: offset variables** - Next by thread:
**st: reduced form** - Index(es):

© Copyright 1996–2024 StataCorp LLC | Terms of use | Privacy | Contact us | What's new | Site index |