[Limdep Nlogit List] Need a kind help

William Greene wgreene at stern.nyu.edu
Fri Apr 7 10:45:24 EST 2006


Dear Mr. Bilgic.  The bivariate tobit model is fit by normalizing each coefficient vector by
the respective standard deviation - the Olsen tranformation.   Unfortunately, this means 
that it is not possible to impose cross equation restrictions in that model.  You can impose
within equation restrictions with ;RST, however.
Sincerely,
B. Greene

************************************************
Professor William Greene
Department of Economics
Stern School of Business
New York University
44 West 4th St., Rm. 7-78
New York, NY   10012
Ph. 212.998.0876
Fax. 212.995.4218
URL. http://www.stern.nyu.edu/~wgreene
Email. wgreene at stern.nyu.edu
************************************************

----- Original Message -----
From: Abdulbaki Bilgic <tebrik at yahoo.com>
Date: Thursday, April 6, 2006 3:11 pm
Subject: Re: [Limdep Nlogit List] Need a kind help

> Dear Dr. Greene,
>  Sorry for this silly mistake and is there any way to impose 
> linear constraints to automated bivariate tobit model such as 
> homogeneity or symmetry (i think symmetry restriction is okay with 
> RST command) restrictions? So far i know there is not (CML does 
> not work under automated bivariate tobit model).
> 
> William Greene <wgreene at stern.nyu.edu> wrote:
>  Mr. Bilgic: I can't vouch for the code for the likelihood function.
> But, you are using 0 and 0 for the starting values for the variances,
> so it's for sure that you cannot compute the function at the starting
> values. That is the source of the problems.
> /B. Greene
> 
> ************************************************
> Professor William Greene
> Department of Economics
> Stern School of Business
> New York University
> 44 West 4th St., Rm. 7-78
> New York, NY 10012
> Ph. 212.998.0876
> Fax. 212.995.4218
> URL. http://www.stern.nyu.edu/~wgreene
> Email. wgreene at stern.nyu.edu
> ************************************************
> 
> ----- Original Message -----
> From: Abdulbaki Bilgic 
> Date: Thursday, April 6, 2006 12:27 pm
> Subject: [Limdep Nlogit List] Need a kind help
> 
> > Dear Limdep Users,
> > I am estimating a bivariate tobit model with three switching 
> > regimes: (y1>0, y2=0), (y1=0,y2>0), and (y1>0, y2>0). The last 
> > term where both y1 and y2 greater than 0 is the conditional 
> > distribution of y2 given y1 >0. I have written a code in Limdep 
> > and got a error reporting a zero division and i have checked the 
> > reference book it says a dependent variable problem and when 
> > deleting the first observation, it gives the same problem too. 
> Any 
> > ideas please.
> > 
> > 
> > CREATE ; IF(Y1 = 0 & Y2 =0)DUMMY1=1
> > ; (ELSE) DUMMY1=0
> > ; IF(Y1 > 0 & Y2 =0)DUMMY2=1
> > ; (ELSE) DUMMY2=0
> > ; IF(Y1 = 0 & Y2 >0)DUMMY3=1
> > ; (ELSE) DUMMY3=0
> > ; IF(Y1 > 0 & Y2 >0)DUMMY4=1
> > ; (ELSE) DUMMY4=0 $
> > NAMELIST ; X1 =ONE
> > ; X2 =ONE $
> > CALC ; LIST
> > ; KA=COL(X1)
> > ; KB=COL(X2)
> > ; KC=KA+1 
> > ; KD=KA+KB $
> > MATRIX ; LIST
> > ; CA=INIT(KA,1,.0)
> > ; CB=INIT(KB,1,.0) $
> > MAXIMIZE ; START =CA,CB,.0,.0,.0
> > ; LABELS=KA_BA, KB_BB,SIGM1,SIGM2,RHO12
> > ; FCN=TA1 =LOG((1+RHO12)/(1-RHO12)) |
> > RHO12 =(EXP(TA1)-1)/(EXP(TA1)+1) |
> > SG1 =1/(SIGM1*(SQR(1-RHO12^2))) | 
> > SG2 =1/(SIGM2*(SQR(1-RHO12^2))) |
> > SG3 =1/(SIGM2^2*(SQR(1-RHO12^2))) |
> > ETA1 =1/(2*SIGM1^2) |
> > ETA2 =1/(2*SIGM2^2) |
> > ETRHO1 =RHO12*(SIGM1/SIGM2) |
> > ETRHO2 =RHO12*(SIGM2/SIGM1) |
> > RHOSQ =(1-RHO12^2) |
> > INDEX1 =BA1'X1 |
> > INDEX2 =BB1'X2 |
> > ERROR1 =Y1-INDEX1 |
> > ERROR2 =Y2-INDEX2 |
> > INDX1 =INDEX1+ETRHO1*ERROR2 |
> > INDX2 =INDEX2+ETRHO2*ERROR1 |
> > INDXX1 =-SG1*INDX1 |
> > INDXX2 =-SG2*INDX2 |
> > DUMMY2*(-0.5*LOG(2*PI)-LOG(SIGM1)-ETA1*ERROR1^2+LOG(PHI(INDXX2)))+
> > DUMMY3*(-0.5*LOG(2*PI)-LOG(SIGM2)-ETA2*ERROR2^2+LOG(PHI(INDXX1)))+
> > DUMMY4*(-0.5*LOG(2*PI)-LOG(SIGM1)-ETA1*ERROR1^2-2*LOG(SIGM2)
> > -0.5*LOG(RHOSQ)-SG3*(Y2-INDX2)^2) 
> > ; ALG=N 
> > ; TLF=1.D-8
> > ; TLG=1.D-4
> > ; SET $
> > 
> > Output is:
> > Error 590: Obs.= 1 Cannot compute function: Zero_dvd
> > Warning 137: Iterations: function not computable at crnt. trial 
> > estimate Error 802: Cannot compute function at start values. 
> Exit 
> > status=4. Function= .36023626764D+05, at entry, .00000000000D+00 
> > at exit
> > +---------------------------------------------+
> > | User Defined Optimization |
> > | Maximum Likelihood Estimates |
> > | Model estimated: Apr 06, 1999 at 04:18:24PM.|
> > | Dependent variable Function |
> > | Weighting variable None |
> > | Number of observations 22344 |
> > | Iterations completed 0 |
> > | Log likelihood function .0000000 |
> > | Number of parameters 0 |
> > | Info. Criterion: AIC = .00000 |
> > | Finite Sample: AIC = .00000 |
> > | Info. Criterion: BIC = .00000 |
> > | Info. Criterion:HQIC = .00000 |
> > +---------------------------------------------+
> > +---------+--------------+----------------+--------+---------+
> > |Variable | Coefficient | Standard Error |b/St.Er.|P[|Z|>z] |
> > +---------+--------------+----------------+--------+---------+
> > BA1 .000000 1.00000000 .000 1.0000
> > BB1 .000000 1.00000000 .000 1.0000
> > SIGM1 .000000 1.00000000 .000 1.0000
> > SIGM2 .000000 1.00000000 .000 1.0000
> > RHO12 .000000 1.00000000 .000 1.0000
> > 
> > Best regards,
> > 
> > 
> > 
> > 
> > ---------------------------------
> > Yahoo! Messenger with Voice. PC-to-Phone calls for ridiculously 
> > low rates.
> > _______________________________________________
> > Limdep site list
> > Limdep at limdep.itls.usyd.edu.au
> > http://limdep.itls.usyd.edu.au
> > 
> 
> _______________________________________________
> Limdep site list
> Limdep at limdep.itls.usyd.edu.au
> http://limdep.itls.usyd.edu.au
> 
> 
>        	
> ---------------------------------
> Yahoo! Messenger with Voice. Make PC-to-Phone Calls to the US (and 
> 30+ countries) for 2¢/min or less.
> _______________________________________________
> Limdep site list
> Limdep at limdep.itls.usyd.edu.au
> http://limdep.itls.usyd.edu.au
> 




More information about the Limdep mailing list