Is a multilevel survival method model better than the basic survival model

[Putu Sintya]

Good day. I'm having a problem of determining whether mestreg model or streg model is more proper for my analysis. A friend of mine telling me that I should look at the result of the LR test and it says that mestreg is not a better model. while the Prob >= chibar2 result says that its ok to use mestreg.
here is my result

Mixed-effects exponential regression Number of obs = 1,608
Group variable: provinsi Number of groups = 33

Obs per group:
min = 18
avg = 48.7
max = 73

Integration method: mvaghermite Integration pts. = 7

Wald chi2(8) = 56.84
Log likelihood = -1245.8049 Prob > chi2 = 0.0000
------------------------------------------------------------------------------
_t | Haz. Ratio Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
jk_anak | 1.52106 .277308 2.30 0.021 1.064049 2.174357
kembar | 4.04597 .9688344 5.84 0.000 2.530445 6.46917
pddk2 | 1.255309 .2566061 1.11 0.266 .8409125 1.873917
Umelahirkan | 1.072102 .0176992 4.22 0.000 1.037968 1.107359
jml_anak | .4816481 .1304414 -2.70 0.007 .2832728 .8189451
UKP | 2.083532 .4368334 3.50 0.000 1.381457 3.142412
rokok | 1.872547 .8757155 1.34 0.180 .7487892 4.682801
desakota | 1.126665 .2177252 0.62 0.537 .7714401 1.64546
_cons | 5.32e-06 3.02e-06 -21.39 0.000 1.75e-06 .0000162
-------------+----------------------------------------------------------------
provinsi |
var(_cons)| .127613 .0991949 .027813 .5855209
------------------------------------------------------------------------------
LR test vs. exponential model: chibar2(01) = 3.09 Prob >= chibar2 = 0.0394

I'm using STATA 14 for Windows

[Steve Samuels]

The last line has p = 0.039, which rejects the hypothesis that the random effect variance is zero. So -mestreg- is better. However, the exponential is a very simple distribution--the mean equals the variance. It might be that another -mestreg- distribution will fit better, with or without a random effect.