Estout: Adding Odds Ratio and Lrtest

[Michael Mazur]

Dear Stata friends,

I have the problem that I do not get the Odds ratio and lrtest information when using the estout command.


melogit teayyopp if mysampleee==1, or || country:, intpoints(30)

Fitting fixed-effects model:

Iteration 0: log likelihood = -286811.83
Iteration 1: log likelihood = -267240.07
Iteration 2: log likelihood = -267210.85
Iteration 3: log likelihood = -267210.85

Refining starting values:

Grid node 0: log likelihood = -260503.05

Fitting full model:

Iteration 0: log likelihood = -260503.05 (not concave)
Iteration 1: log likelihood = -260500.31 (not concave)
Iteration 2: log likelihood = -260497.57 (not concave)
Iteration 3: log likelihood = -260495.25 (not concave)
Iteration 4: log likelihood = -260493.33 (not concave)
Iteration 5: log likelihood = -260491.76 (not concave)
Iteration 6: log likelihood = -260490.52
Iteration 7: log likelihood = -260489.54
Iteration 8: log likelihood = -260488.66
Iteration 9: log likelihood = -260488.52
Iteration 10: log likelihood = -260488.49
Iteration 11: log likelihood = -260488.49

Mixed-effects logistic regression Number of obs = 1277158
Group variable: country Number of groups = 33

Obs per group:
min = 8,119
avg = 38,701.8
max = 315,334

Integration method: mvaghermite Integration pts. = 30

Wald chi2(0) = .
Log likelihood = -260488.49 Prob > chi2 = .
------------------------------------------------------------------------------
teayyopp | Odds Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
_cons | .0601004 .004484 -37.69 0.000 .0519242 .0695639
-------------+----------------------------------------------------------------
country |
var(_cons)| .1827738 .0451973 .1125704 .296759
------------------------------------------------------------------------------
Note: Estimates are transformed only in the first equation.
LR test vs. logistic model: chibar2(01) = 13444.71 Prob >= chibar2 = 0.0000

. estimates store model2

estout * using model, replace cells((b(fmt(3) label("b_Coeff")) p(par fmt(3) label("p_Val")) _star(label("Star")) se(fmt(3) label("Std_Err")))) starlevels(* 0.10 ** 0.05 *** 0.01) stats(N F chi2 df df_m aic bic lrtest_chi2 lrtest_df lrtest_p)

Output:

	model2
	b_Coeff
teayyopp
_cons	-2.812
/
var(_cons[country])	0.183
N	1.277.158.000
F
chi2
df
df_m	0.000
aic	520.980.982
bic	521.005.102
lrtest_chi2
lrtest_df
lrtest_p

The output shows the b coefficient and nothing for the lrtest results. However, you can see from above the lrtest is 13444.71 and p-value is 0.0000

Do you know the reason for that issue?

Thank you very much!

Best wishes,
Michael

[Roman Mostazir]

Use 'eform' option after 'estout' command to get the results back transformed. For stats option, you are using several names that do not exist in the scalars after 'melogit' such as F. AIC, BIC arre calculated by 'estout' but you need to run 'ereturn' after the model to see what are the available scalars estout can print for you in the stat option.

melogit your_model

est sto your_model_name

eret li // see the available e() scalars for estout can print

estout your_model_name, eform stats(N p chi2 chi2_c ll_c) //some available e() scalars after melogit

PS: Please learn to use code delimiters for posting Stata outputs here. Read the forum rules FAQ section. The way you posted your output is quite difficult to understand.

[Michael Mazur]

Thank you very much!!!!

It worked.

Is it also possible with the estout command to add number of groups (here countries), Prob > Chibar2 test, ICC and IC (AIC, BIC). I did not find it in the eret li list.

Regarding posting Stata outputs - I will keep this in mind. Sorry for that!

I highly appreciate your help!

Best wishes,
Michael

[Roman Mostazir]

That information is saved in a matrix e(N_g). You need to work around a little to add that in your result using 'estadd' , see below:

est sto your_model_name

mat  r = e(N_g)

estadd loc Group = r[1,1], replace

estout your_model_name, eform stats(N  Group p chi2 chi2_c ll_c )

[Michael Mazur]

Thank you! It worked

Best regards,
Michael