How do I use NLIN procedures' Hougaard skewness and Box' Bias measures in SAS MODEL procedure?

[iskenderc]

I have been doing a multiple regression (nonlinear function) application with lag in SAS MODEL procedure. At the same, time I also want to apply the Hougaard's skewness and Box's bias tests of NLIN procedure for this application. There is no statement for lagged variable in NLIN. I'm waiting for the help of experts. Thanks.

[sbxkoenk]

Hello,

Hougaard’s skewness measure and Box’s bias measure are not available in PROC MODEL (nor in PROC TSMODEL as far as I know).

And replicating your PROC MODEL analysis in PROC NLIN will probably not be successful. You can of course use lagged variables in NLIN (if you do some data prep in data step), but you do not have the dynamic modeling capabilities and special parameter estimation methods that PROC (TS)MODEL has.

Can you show us your PROC MODEL code?? But again, little chance your analysis can be replicated in PROC NLIN.

Thanks,

Koen

[sbxkoenk]

Hello,

Hougaard’s skewness measure and Box’s bias measure are not available in PROC MODEL (nor in PROC TSMODEL as far as I know).

And replicating your PROC MODEL analysis in PROC NLIN will probably not be successful. You can of course use lagged variables in NLIN (if you do some data prep in data step), but you do not have the dynamic modeling capabilities and special parameter estimation methods that PROC (TS)MODEL has. Nor can you have simultaneous systems of nonlinear equations in PROC NLIN, I think.

Can you show us your PROC MODEL code?? But again, little chance that you can replicate your analysis in PROC NLIN.

Thanks,

Koen

[iskenderc]

Thanks for your very valuable helps:

Here is my PROC MODEL code:

--------------------------------------------

ods graphics on;
PROC MODEL data=bugday plots=all ;
PARAMETERS L=2500000 re=-2.3 Qee=0.27 Ke=16.71 ;
T=1940;
A1=-700000 ;
c1=10;
d1=7.587 ;
b1=0;
Q=EXP(EXP(Qee));
r=EXP(re);
K=EXP(Ke);
E=(1+Q*EXP(-r*(Year-T)));
R1=(A1)*EXP(b1*(year-t))*cos(c1*(year-t)+d1);
Wheat=L+R1+((K-L)/E);
z=year-1950;
en=year-1990;
s=en*z;
%ma(wheat,2,M=ML);
fit Wheat / out=sdf2 FIML fsrsq prl=both covs covbest= cross
details dw dwprob corr collin white pagan=(1 s wheat)
normaL printall GODFREY=3 dW=4 list outall
chow =(40 60 80) pchow=90;
RUN;

--------------------------

[iskenderc · Accepted solution]

Finally,

I calculated lags of MODEL procedures then Applied them to NLIN procedure. This was indirect method but successful.

Thanks for helps.