Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


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

Re: st: RE: GLLAMM: logitic regression interaction non signficant but lincom gives signficant result


From   David Hoaglin <dchoaglin@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: RE: GLLAMM: logitic regression interaction non signficant but lincom gives signficant result
Date   Sat, 19 May 2012 16:35:32 -0400

Abdelouahid,

You are confused because you see a contradiction where there is none.
A similar situation can arise in comparing the means of several groups
in a one-way analysis of variance.  The F-test can reject the
hypothesis that all the true means are equal; but, when the means are
put in increasing order, all the comparisons between adjacent means
can be non-significant (allowing appropriately for the multiple
comparisons).

David Hoaglin

On Sat, May 19, 2012 at 1:22 PM, Abdelouahid Tajar
<a_tajar@hotmail.co.uk> wrote:
> Thanks for all the comments and the references,
>
> Indeed the interaction term tests if the two slopes are significantly different from each other and you only get the slope and its confidence interval for the reference group, however lincom " computes point estimates, standard errors, t or z statistics, p-values, and confidence intervals for linear combinations of coefficients after any estimation command".
>
> Because I wanted to compute the slope of group1 I run the lincom, BUT, the lincom result are in contradiction ( in terms of significance) with the interaction  result in the main model. The main model says that the two slopes are not significantly different (interaction non significant0 and because the slope of the reference is not significantly different from 0 this should mean that the slope of group1 is also not significantly different from 0 but the lincom result shows that the slope of group 1 is actually significantly different from 0, hence my confusion.
>
> Denominator in gllamm is for the number of trials, hence for a Bernoulli distribution denominator is set to be 1.
>  See Page 374 in S. Rabe-Hesketh and A Skrondal: Multilevel and longitudinal modeling using stata.
>
> I think the contradiction between the non significance result in the interaction term and the significance result in the lincom is not necessarily related to gllamm.

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index