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

From |
"Wendell Joice" <wjoice@erols.com> |

To |
<statalist@hsphsun2.harvard.edu> |

Subject |
st: lincom and coefficients |

Date |
Fri, 4 Apr 2003 11:19:16 -0500 |

Dear Statalist, Using Stata 7, I am trying to use -lincom- to test some partial interaction effects after xi:logistic regression. I would like to compare one of the partial interactions with another that is part of the reference group. I can't quite figure it out. In fact, is there a simple explanation on how to use -lincom- to test interaction factors that are partially obscured by being in the reference group. The example in the manual uses smokers and race: the reference group is White nonsmokers. When you want to test a factor that is in the reference group, say White smokers, it appears that you use -Black. How does that work? More specifically, in my case: Example: xi: logistic psyment i.racewbh*i.agecatc i.sex i.met i.patseen i.psydin2 i.yearmhp if csample==1, robust cluster(phycoden) psyment | Odds Ratio Std. Err. z P>|z| [95% Conf. Interval] -------------+-------------------------------------------------------- -------- _Iracewbh_2 | .5307681 .1412873 -2.38 0.017 .3150069 .8943132 _Iracewbh_3 | .6093152 .1890877 -1.60 0.110 .3316578 1.119422 _Iagecatc_2 | 1.291446 .1788368 1.85 0.065 .9844729 1.694139 _Iagecatc_3 | 1.179668 .1588651 1.23 0.220 .9060019 1.535998 _Iagecatc_4 | .6028866 .0872337 -3.50 0.000 .4540172 .8005692 _IracXag~2_2 | 1.543127 .4586715 1.46 0.144 .8617767 2.763177 _IracXag~2_3 | 1.017071 .3300086 0.05 0.958 .5384711 1.921057 _IracXag~2_4 | 1.295049 .4516837 0.74 0.459 .6537424 2.565465 _IracXag~3_2 | 1.226491 .4026047 0.62 0.534 .6445417 2.333875 _IracXag~3_3 | 1.334677 .4433691 0.87 0.385 .696007 2.559402 _IracXag~3_4 | 1.24216 .504893 0.53 0.594 .5600092 2.755245 etc etc For example, I want to test _IracXag~2_3_ vs _IracXag~1_3, and _IracXag~3_3 vs _IracXag~1_3, etc.... Since _IracXag~1_3 is partially part of the reference group, how do I make these tests using -lincom-? Is this the best approach for such tests? Should I be using another coding system via xi2 ...or will that just be extra strain on the brain with minimal gain? ...smile * * 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/

- Prev by Date:
**st: RE: Displaying text without displaying the display text** - Next by Date:
**Re: st: RE: fill bars** - Previous by thread:
**st: Displaying text without displaying the display text** - Next by thread:
**st: Determining the width of a formatted value** - Index(es):

© Copyright 1996–2016 StataCorp LP | Terms of use | Privacy | Contact us | What's new | Site index |