# No Value for Wald chi2 or prob with random effect in negative binomial regression

#### atwell17

##### New Member
Hello all! So I'm doing a negative binomial regression, for my DV (timecz) and the interaction of two IVs (predation and age). I added in a random factor, spkr, and I get no values for the Wald chi2 or prob. 'spkr' has values 1, 2, 3, 4 so it is a categorical variable, and every one of my data points has a 'spkr' value to it.

I've also used a different random factor, family, and I get the Wald chi2 and prob values. 'family' is what family the subject came from, so again it is categorical and every data point has a value, but there are about 20 different family values.

I would appreciate any insight into this problem. Thank you!

menbreg timecz predation##age || spkr:, vce(robust)

Fitting fixed-effects model:

Iteration 0: log likelihood = -610.48561
Iteration 1: log likelihood = -585.81128
Iteration 2: log likelihood = -584.98962
Iteration 3: log likelihood = -584.9823
Iteration 4: log likelihood = -584.9823

Refining starting values:

Grid node 0: log likelihood = -586.93141

Fitting full model:

Iteration 0: log pseudolikelihood = -586.93141 (not concave)
Iteration 1: log pseudolikelihood = -584.50204
Iteration 2: log pseudolikelihood = -584.30355
Iteration 3: log pseudolikelihood = -584.24794
Iteration 4: log pseudolikelihood = -584.24786
Iteration 5: log pseudolikelihood = -584.24786

Mixed-effects nbinomial regression Number of obs = 117
Overdispersion: mean
Group variable: spkr Number of groups = 4

Obs per group: min = 27
avg = 29.2
max = 32

Integration method: mvaghermite Integration points = 7

Wald chi2(3) = .
Log pseudolikelihood = -584.24786 Prob > chi2 = .
(Std. Err. adjusted for clustering on spkr)
-------------------------------------------------------------------------------
| Robust
timecz | Coef. Std. Err. z P>|z| [95% Conf. Interval]
--------------+----------------------------------------------------------------
|
2.predation | .0265594 .6956106 0.04 0.970 -1.336812 1.389931
|
age |
2 | .4622833 .2973759 1.55 0.120 -.1205628 1.045129
3 | .5894719 .5789841 1.02 0.309 -.545316 1.72426
|
predation#age |
2 2 | .6903936 .5020987 1.38 0.169 -.2937017 1.674489
2 3 | .0129458 1.045617 0.01 0.990 -2.036425 2.062317
|
_cons | 3.627404 .3742223 9.69 0.000 2.893942 4.360866
--------------+----------------------------------------------------------------
/lnalpha | .6274604 .1765955 3.55 0.000 .2813397 .9735812
--------------+----------------------------------------------------------------
spkr |
var(_cons)| .0659578 .0592337 .0113459 .3834348
-------------------------------------------------------------------------------

#### atwell17

##### New Member
If anyone comes across this and wondered what happened, I ended up going with a zero inflated negative binomial model instead due to the much better fit, so this wasn't an issue anymore.