Notice: On March 31, it was **announced** that Statalist is moving from an email list to a **forum**. The old list will shut down on April 23, and its replacement, **statalist.org** is already up and running.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

From |
David Hoaglin <dchoaglin@gmail.com> |

To |
statalist@hsphsun2.harvard.edu |

Subject |
Re: st: time in xtmixed |

Date |
Mon, 25 Mar 2013 15:33:20 -0400 |

Megan, Those results are a bit unusual. They suggest a need for a close look at the data. (I have not seen information on the number of observations or the nature of depvar or indepvar or, as mentioned in another comment, the units of time.) You might, for example, plot depvar vs. indepvar separately for each of the three time points, and also look at how indepvar is related to time. My rough interpretation is that the contribution of time is not linear, so specifying time as categorical adjusts more effectively for that contribution and leaves less for indepvar to account for. It is not out of the question (obviously, since you see it in your data) for time to have a (nearly) significant trend without either the time1 or the time2 effect being significant. Those effects reflect the contribution of time to depvar after adjusting for the contribution of indepvar, so examination of the relation between indepvar and time is likely to be important. Have you changed your models? The models in your first message did not include an interaction between indepvar and time. I'm not sure what you mean by "the interaction between indepvar and time ... at time 0." Even if your models include an interaction between indepvar and time, the contribution at time 0 would be part of the constant term (unless you center time). If you went with option a, it would not be appropriate to say that, independent of time, depvar and indepvar are associated. The appropriate statement would be that, after adjusting for the linear effect of time, depvar has a significant slope against indepvar. The suggestion that the contribution of time is not linear, however, may mean that even that more-careful statement is not a good summary of your data. Since you have only three time points, it will probably not be useful to consider polynomials in time. Linear time is already the first-order polynomial, and a quadratic in time would fit three time points exactly. David Hoaglin On Mon, Mar 25, 2013 at 9:12 AM, megan rossi <megan_rossi@msn.com> wrote: > > How is this for confusing, when time is as a continuous variable (ie. option a)the association between depvar and indepvar is significant (time is borderline sig p=0.055), however as a categorical variable, option b, the relationship between depvar and indepvar becomes insignificant...and time 1 and 2 become very insignificant (p=0.55 and 0.17) > In both scenario's the overall significance of the model is <0.0001 and log likelihoods are the same. The interaction between indepvar and time is only significant at time 0 not 1 or 2. > If I did go with option a, could I really say that independent of time depvar and indepvar are association? > Megan Rossi APD * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/faqs/resources/statalist-faq/ * http://www.ats.ucla.edu/stat/stata/

**References**:**st: time in xtmixed***From:*megan rossi <megan_rossi@msn.com>

**Re: st: time in xtmixed***From:*David Hoaglin <dchoaglin@gmail.com>

**st: time in xtmixed***From:*megan rossi <megan_rossi@msn.com>

- Prev by Date:
**st: IVs for logit models using panel data** - Next by Date:
**Re: st: Generalized Lorenz Curves with Confidence Intervals** - Previous by thread:
**Re: st: time in xtmixed** - Next by thread:
**st: time in xtmixed** - Index(es):