topic Re: Panel Data Non-Linear Model in Statistical Procedures

Panel Data Non-Linear Model

Sadia_Sharmin — Fri, 29 Jul 2016 08:05:32 GMT

I have a panel data set with overdispersion. I want to run Fixed Effect and Random Effect Negative Binomial distribution model. I have SAS 9.4.

proc countreg data=CS.total groupid= loc;
class rtl(ref='0')majl (ref='2')minl (ref='2') ydr1 (ref='0') lftl (ref='0')/ param=ref;
model tot = majl minl lftl rtl majadt minadt ydr dui /dist=negbin errorcomp=fixed;
run;

SAS cannot recognize groupid, dist, errorcomp terms, as they didnt appear blue. How to code for my data structure?

Thanks in Advance

Re: Panel Data Non-Linear Model

Reeza — Fri, 29 Jul 2016 08:12:52 GMT

Colour isn't indicative of correct code. Sadly the syntax highlighter isn't always correct.

Did you get any errors? Are any of your statements not listed in the procedure documentation?

Re: Panel Data Non-Linear Model

Sadia_Sharmin — Fri, 29 Jul 2016 08:22:21 GMT

Thank you for your reply dear Reeza.

Error: NEWAP Optimization cannot be completed.

WARNING: The Hessian matrix is singular.

And I am not sure about my coding as couldnot find any example using this code. And my data structure is like :

data CS.total;
input loc t majl $ minl $ lftl $ rtl $ majadt minadt tev ydr ydr1 $ mdr odr dui tot FI PDO angle re ss;
datalines;
1 1 4 4 1 1 9.8020 8.8465 10.1274 4.9940 1 20.4509 0.0000 0.1429 7 3 4 2 2 0
1 2 4 4 1 1 9.7791 8.8235 10.1045 13.3351 3 9.6237 0.0000 0.0000 10 0 10 2 3 2
1 3 4 4 1 1 9.7886 8.8331 10.1141 8.3439 2 27.6996 0.0000 0.0909 11 5 6 6 2 0
1 4 4 4 1 1 9.8215 8.8660 10.1469 15.1548 4 17.6103 0.0000 0.0769 13 2 11 6 4 2.....

I have 15 loc variables 1-15 and time series variable for 21 years, so come characteristics reapeted over the same years but I have also random variables. I want to compare FENB, RENB and NB models for my data. Can you please help me out with SAS code which will work in SAS 9.4 version? And I am really confused about my coding.

Re: Panel Data Non-Linear Model

Sadia_Sharmin — Fri, 29 Jul 2016 08:52:57 GMT

474 *Considering AADT/;
475 *-- FE Neg Bin Model --*/;

476 proc countreg data=CS.total groupid= loc;
NOTE: Writing HTML Body file: sashtml.htm
477 class rtl(ref='0')majl (ref='2')minl (ref='2') ydr1 (ref='0') lftl
477! (ref='0')/ param=ref;
478 model tot = majl minl lftl rtl majadt minadt ydr dui /dist=negbin
478! errorcomp=fixed;
479 run;

ERROR: NEWRAP Optimization cannot be completed.
NOTE: Optimization routine cannot improve the function value.
WARNING: The Hessian matrix is singular.

This is the programing log. From here I think ref for my categorical variables and errorcomp is not recognized by SAS.

TIA

Re: Panel Data Non-Linear Model

Ksharp — Sat, 30 Jul 2016 05:33:06 GMT

Since it is a time series problem (SAS/ETS) , I suggest you post it at another forum:
Communities
SAS Forecasting and Econometrics