How to interpret LR test in xtmelogit??

[jaewon kim]

Below is my command and result.

. xtmelogit dic year##newb##sole age sex marry edu_g2 f_size home wealth hh_incom p_pn pr_trans econpart2 house soc_r fam_r c_dx inhc || hhid: || id:

Refining starting values:

Iteration 0: log likelihood = -923.00221
Iteration 1: log likelihood = -917.17197
Iteration 2: log likelihood = -916.79587

Performing gradient-based optimization:

Iteration 0: log likelihood = -916.79587
Iteration 1: log likelihood = -916.7067
Iteration 2: log likelihood = -916.70603
Iteration 3: log likelihood = -916.70602

Mixed-effects logistic regression Number of obs = 1,520

----------------------------------------------------------------------------
| No. of Observations per Group Integration
Group Variable | Groups Minimum Average Maximum Points
----------------+-----------------------------------------------------------
hhid | 620 2 2.5 4 7
id | 760 2 2.0 2 7
----------------------------------------------------------------------------

Wald chi2(23) = 122.92
Log likelihood = -916.70602 Prob > chi2 = 0.0000

--------------------------------------------------------------------------------
dic | Coef. Std. Err. z P>|z| [95% Conf. Interval]
---------------+----------------------------------------------------------------
year |
2008 | -.2730442 .2375819 -1.15 0.250 -.7386962 .1926078
|
1.newb | -.1211041 .3129108 -0.39 0.699 -.734398 .4921898
|
year#newb |
2008 1 | .3815499 .3574489 1.07 0.286 -.3190371 1.082137
|
1.sole | -.4997881 .3289285 -1.52 0.129 -1.144476 .1449
|
year#sole |
2008 1 | .2052514 .3413537 0.60 0.548 -.4637896 .8742924
|
newb#sole |
1 1 | .5648403 .4266411 1.32 0.186 -.2713609 1.401042
|
year#newb#sole |
2008 1 1 | -.9253963 .5023913 -1.84 0.065 -1.910065 .0592726
|
age | -.0220127 .030512 -0.72 0.471 -.0818152 .0377898
sex | .2641683 .1906669 1.39 0.166 -.1095319 .6378685
marry | .56805 .3670882 1.55 0.122 -.1514297 1.28753
edu_g2 | -.2518682 .1399016 -1.80 0.072 -.5260703 .022334
f_size | .0081469 .0967724 0.08 0.933 -.1815236 .1978174
home | .3902715 .3192878 1.22 0.222 -.235521 1.016064
wealth | -.0010083 .0004767 -2.12 0.034 -.0019425 -.000074
hh_incom | -.0072073 .0061165 -1.18 0.239 -.0191955 .0047809
p_pn | -.0003896 .0001501 -2.60 0.009 -.0006838 -.0000954
pr_trans | .0000963 .0001529 0.63 0.529 -.0002033 .0003959
econpart2 | -.0624379 .1811537 -0.34 0.730 -.4174925 .2926168
house | .2714114 .2234762 1.21 0.225 -.1665939 .7094168
soc_r | -.3065462 .1520254 -2.02 0.044 -.6045106 -.0085818
fam_r | -.5668335 .168877 -3.36 0.001 -.8978263 -.2358407
c_dx | .5227505 .2146773 2.44 0.015 .1019907 .9435103
inhc | .6362042 .1810271 3.51 0.000 .2813976 .9910109
_cons | 1.869337 2.312023 0.81 0.419 -2.662145 6.40082
--------------------------------------------------------------------------------

------------------------------------------------------------------------------
Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval]
-----------------------------+------------------------------------------------
hhid: Identity |
sd(_cons) | 1.118888 .1724824 .827122 1.513574
-----------------------------+------------------------------------------------
id: Identity |
sd(_cons) | .3130409 .5917043 .0077033 12.72117
------------------------------------------------------------------------------
LR test vs. logistic model: chi2(2) = 38.10 Prob > chi2 = 0.0000

Note: LR test is conservative and provided only for reference.


================================================== ========

And I know for mixed model, we have to do LR test to decide Random Intercept or Random Slope model.
But my academic interest is get one beta, not to see the variation in population
so I didn't do that.

However, in the bottom of the result, there is

------------------------------------------------------------------------------
LR test vs. logistic model: chi2(2) = 38.10 Prob > chi2 = 0.0000

Note: LR test is conservative and provided only for reference.


I want to interpret this part, but I cannot find proper reference....

Can anybody give me answer.................?

Thanks in advance!

[Carlo Lazzaro]

Jaewon:
I'm not clear about your concern.
You were seemingly performing a three-level intercept-only mixed model: this topic is covered in Example 4, -xtmelogit- entry, Stata 12.1 .pdf manual (I assume - and you should have declared it yourself - that you refer to that Stata release, as from Stata 13 onwards -xtmelogit- has been superseded by -meqrlogit-).
As an aside, for the future please post what you typed and what Stata gave you back via CODE delimiters (see the FAQ on that topic), as your Stata output has serious formatting issues. Thanks.

[jaewon kim]

Carlo Lazzaro

What I wonder is just to know the meaning of the LR test. a.k.a. to interpret the LR test at the bottom of result.
^^

[Carlo Lazzaro]

Jaewon:
LR test tells you that -xtmelogit- outperforms logistic regression. Hence, go -xtmelogit-.

[Marcos Almeida]

Carlo already clarified the interpretation of the LR test. That said, please keep in mind it is usually expected, since your data is under a multi-level structure.

To end, and assuming you are using Stata 14, this is the information we get when typing xtmelogit:

xtmelogit has been renamed to meqrlogit. xtmelogit continues to work but, as of Stata 13, is no longer an official part of Stata. This is the original help file, which we will no longer update, so some links may no longer work.

[jaewon kim]

Thank you, Carlo Lazzaro , Marcos Almeida