"matrix operation not found"- Tobit MLE Spatial Error Panel Normal Model (SEM)

[Yan Li]

Good day

I am using conmand "sptobitsemxt " to run the Tobit MLE Spatial Error Panel Normal Model (SEM). (Ps. my dataset is a banlanced panel with 31 provinces during 2012 to 2016)

However, "matrix operation not found" appears in the results.(And when I run "sptobitsdmxt" for the Tobit MLE Spatial Panel Durbin Regression, the same problem appears again.)

My command is below:

sptobitsemxt te_cm gra urb ip inv fd ipr c inc,nc(31) wmfile(W.dta) dist(norm) ll(0)



The results are as follows:



================================================== ============================
*** Binary (0/1) Weight Matrix: 155x155 - NC=31 NT=5 (Non Normalized)
================================================== ============================
*** te_cm Lower Limit = 0
*** te_cm Left- Censored Observations = 0
*** te_cm Left-UnCensored Observations = 155
------------------------------------------------------------

initial: log likelihood = 67.159033
rescale: log likelihood = 67.159033
rescale eq: log likelihood = 67.159033
Iteration 0: log likelihood = 67.159033
Iteration 1: log likelihood = 68.597513
Iteration 2: log likelihood = 69.101243
Iteration 3: log likelihood = 70.055014
Iteration 4: log likelihood = 70.069662
Iteration 5: log likelihood = 70.069683
Iteration 6: log likelihood = 70.069683
================================================== ============================
* Tobit MLE Spatial Error Panel Normal Model (SEM)
================================================== ============================
te_cm = gra + urb + ip + inv + fd + ipr + c + inc
------------------------------------------------------------------------------
Sample Size = 155 | Cross Sections Number = 31
Wald Test = 364.1287 | P-Value > Chi2(8) = 0.0000
F-Test = 45.5161 | P-Value > F(8 , 116) = 0.0000
(Buse 1973) R2 = 0.7138 | Raw Moments R2 = 0.8562
(Buse 1973) R2 Adj = 0.6200 | Raw Moments R2 Adj = 0.8090
Root MSE (Sigma) = 0.1534 | Log Likelihood Function = 70.0697
------------------------------------------------------------------------------
- R2h= 0.6000 R2h Adj= 0.4689 F-Test = 27.37 P-Value > F(8 , 116) 0.0000
- R2v= 0.6050 R2v Adj= 0.4756 F-Test = 27.95 P-Value > F(8 , 116) 0.0000
------------------------------------------------------------------------------
| Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
te_cm |
gra | -.0362367 .0158576 -2.29 0.022 -.067317 -.0051564
urb | -.1200604 .1230545 -0.98 0.329 -.3612428 .121122
ip | .0004303 .0145258 0.03 0.976 -.0280398 .0289004
inv | .0435578 .0247127 1.76 0.078 -.0048782 .0919938
fd | -.0068297 .0016547 -4.13 0.000 -.0100728 -.0035866
ipr | .0098452 .001361 7.23 0.000 .0071776 .0125127
c | .0229674 .0342389 0.67 0.502 -.0441396 .0900744
inc | .2329174 .0820685 2.84 0.005 .072066 .3937687
_cons | -3.101881 .7988903 -3.88 0.000 -4.667678 -1.536085
-------------+----------------------------------------------------------------
/Lambda | -.0056261 .0026722 -2.11 0.035 -.0108634 -.0003887
/Sigma | .1539588 .0087443 17.61 0.000 .1368204 .1710973
------------------------------------------------------------------------------
LR Test SEM vs. OLS (Lambda=0): 4.4329 P-Value > Chi2(1) 0.0353
Acceptable Range for Lambda: -0.3692 < Lambda < 0.1894
------------------------------------------------------------------------------
matrix operation not found
r(501);

end of do-file

r(501);


Very much looking for your kind advice, and massive thanksss in advance.

Best regards,

Yan

[Phil Bromiley]

You'll increase your chances of a helpful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output (fixed spacing fonts help), and sample data using dataex.

You are using a user-written procedure. You might check if a Stata provided procedure will handle your problem - Stata 15 has more spatial models than earlier versions.

Unless someone on the listserve happens to use that procedure and happens to have had the problem, you are not likely to get a good answer. It is conceivable that the procedure uses an obsolete matrix specification. You can either try to dig into the ado file for the procedure to diagnose the problem yourself, or you can contact the authors.

[Yan Li]

Phil Bromiley Thank you so much for your kind suggestion. Regarding to spatial models, I noticed that Stata 15 provides new procedure to deal with the continuous variables (i.e., spregress). However, as for the censored data, I only see the procedures provided by prof.Shehata (https://www.statalist.org/forums/for...ression-models). Also I have written to Prof. Shehata and waiting for his response.

Would you please kindly recommend me some clues for the new spatial Tobit models (procedures) in Stata 15 if I ignored any of them?

A massive thanks in advance.

Best regards.

[Valeriya Lakshina]

Hello everyone. I have the same issue with program sptobitsdmxt. I added -set trace on- in ado file and found out that the error is thrown in line 575 of sptobitsdmxt.ado:

matrix `b'=e(b)

In order to reproduce the error it's possible to use code from https://ideas.repec.org/c/boc/bocode/s457694.html
I attached it to the post for the convenience. The error appears after running line 6 from sptobitsdmxt.do:

sptobitsdmxt ys x1 x2 , nc(7) wmfile(SPWxt) mfx(lin) test ll(0) dist(norm)

Thanks in advance for your answers.

Kind regards,
Valeriya

P.S. Sorry for the formatting, my first post.