I am using a mixed model regression to determine what influences my response variable (V1), including an explanatory variable (T1) and 7 point splines fit to Julian day (spline*) and time of day (pline*). I need to account for random effects of individual nested within Julian day, as well as auto correlation.


Code:
* Example generated by -dataex-. To install: ssc install dataex
clear
input double(AKST_datetime stat_id) float JD double(V1 T1) float(spline1 spline2 spline3 spline4 spline5 spline6 pline1 pline2 pline3 pline4 pline5 pline6)
1.8091296e+12 4001 120 38.541666666666664 6.3416666984558105 120 0 0 0 0 0 0           0 0 0 0 0
1.8091296e+12 4003 120                 38 6.3416666984558105 120 0 0 0 0 0 0           0 0 0 0 0
1.8091296e+12 4006 120            37.9375 6.3416666984558105 120 0 0 0 0 0 0           0 0 0 0 0
1.8091296e+12 4010 120                 38 6.3416666984558105 120 0 0 0 0 0 0           0 0 0 0 0
1.8091296e+12 4011 120 37.541666666666664 6.3416666984558105 120 0 0 0 0 0 0           0 0 0 0 0
1.8091296e+12 4012 120 38.479166666666664 6.3416666984558105 120 0 0 0 0 0 0           0 0 0 0 0
1.8091296e+12 4013 120              37.75 6.3416666984558105 120 0 0 0 0 0 0           0 0 0 0 0
1.8091332e+12 4001 120               38.5  4.816666603088379 120 0 0 0 0 0 1 .0018903592 0 0 0 0
1.8091332e+12 4003 120                 38  4.816666603088379 120 0 0 0 0 0 1 .0018903592 0 0 0 0
1.8091332e+12 4006 120 37.583333333333336  4.816666603088379 120 0 0 0 0 0 1 .0018903592 0 0 0 0
1.8091332e+12 4010 120 38.020833333333336  4.816666603088379 120 0 0 0 0 0 1 .0018903592 0 0 0 0
1.8091332e+12 4011 120 37.666666666666664  4.816666603088379 120 0 0 0 0 0 1 .0018903592 0 0 0 0
1.8091332e+12 4012 120 38.458333333333336  4.816666603088379 120 0 0 0 0 0 1 .0018903592 0 0 0 0
1.8091332e+12 4013 120              37.75  4.816666603088379 120 0 0 0 0 0 1 .0018903592 0 0 0 0
1.8091368e+12 4001 120               38.5  4.083333492279053 120 0 0 0 0 0 2  .015122874 0 0 0 0
1.8091368e+12 4003 120                 38  4.083333492279053 120 0 0 0 0 0 2  .015122874 0 0 0 0
1.8091368e+12 4006 120            37.5625  4.083333492279053 120 0 0 0 0 0 2  .015122874 0 0 0 0
1.8091368e+12 4010 120                 38  4.083333492279053 120 0 0 0 0 0 2  .015122874 0 0 0 0
1.8091368e+12 4011 120 37.708333333333336  4.083333492279053 120 0 0 0 0 0 2  .015122874 0 0 0 0
1.8091368e+12 4012 120            38.3125  4.083333492279053 120 0 0 0 0 0 2  .015122874 0 0 0 0
1.8091368e+12 4013 120              37.75  4.083333492279053 120 0 0 0 0 0 2  .015122874 0 0 0 0
1.8091404e+12 4001 120            37.9375 3.6083333492279053 120 0 0 0 0 0 3    .0510397 0 0 0 0
1.8091404e+12 4003 120 37.979166666666664 3.6083333492279053 120 0 0 0 0 0 3    .0510397 0 0 0 0
1.8091404e+12 4006 120 37.729166666666664 3.6083333492279053 120 0 0 0 0 0 3    .0510397 0 0 0 0
1.8091404e+12 4010 120                 38 3.6083333492279053 120 0 0 0 0 0 3    .0510397 0 0 0 0
1.8091404e+12 4011 120               37.5 3.6083333492279053 120 0 0 0 0 0 3    .0510397 0 0 0 0
1.8091404e+12 4012 120              38.25 3.6083333492279053 120 0 0 0 0 0 3    .0510397 0 0 0 0
1.8091404e+12 4013 120              37.75 3.6083333492279053 120 0 0 0 0 0 3    .0510397 0 0 0 0
end
format %tc AKST_datetime
Below is the model I am trying to run

mixed V1 T1 spline* pline*||stat_id:||c.JD:, residuals(ar, t(AKST_datetime) by(stat_id))

Unfortunately, when I run this model, or even the null model without any explanatory variables, I get the following errors:

I(): 3900 unable to allocate real <tmp>[82800001,82800001]
_xtm_get_irgamma(): - function returned error
_xtm_mult_resid(): - function returned error
_xtm_mixed_ll_uu(): - function returned error
_xtm_mixed_ll_u(): - function returned error
_xtm_mixed_ll(): - function returned error
<istmt>: - function returned error

I have paired this data set down to the bare minimum and still get the same error. I am at a loss to why this is not working as it works just fine with my other data sets.

Any insight would be greatly appreciated.

Stata SE 15.1 with latest update

Thanks!
Dan Thompson