The RATS Software Forum

Hi Tom, thanks for your clarification.. pls, i got this result, the signs are ok but the P-value are all not signif.. can i go ahead and kalman filter with these estimates.. DLM - Estimation by BFGS Convergence in 20 Iterations. Final criterion was 0.0000078 <= 0.0000100 Quarterly Data From 1970:01 ...

Thanks Tom.. when i pegged the Eta to zero, it implies that the parameter is now constant.. is it the same as removing the parameter from the state vector? i mean makinf that particular parameter part of the observable.
Thanks

Hello Tom, I am sorry to bother you again.. please could you advise on what to do to get positive value for Eta in the estimation below. I have tried severally to no avail.. Nonlin d3 sigma eta psi zeta compute sigma =initsig*0.00009,eta =initsig*0.000001 compute psi =initsig*0.001, zeta = initsig*0...

Hi Tom,
Thanks very much..

Hi, I performed a rolling regression of the form: compute start = 1970:1, end = 1985:4 do i=0,96 linreg lexp start+i end+i # constant ltwm lrp lexp{1} end Can someone please tell me how to graph the Betas in the rolling regression.. I want to generate a graph like the one in the paper below: "K...

Hi Tom,
Just wanna say thanks to you for all your assistance..
I have been able to make alot of progress as a result of your consistent assistance..
Thanks once again

Hi Tom, Thanks very much for all the assistance.. I am making progress with my work, thanks be to u.. meanwhile, i have 3 questions.. first, is it proper for theta, when concentrating out one of the variance to be negative? secondly, i have 3 components in the state vector, B0, B1 and B2. but B0 is ...

Hi Tom, please i got these results from my estimation.. the first was done when i estimate the data in their level, but the result was obtain when i multiply the data set by 100 prior to the estimation.. which of them will favourably be consider? (1) DLM - Estimation by BFGS Convergence in 58 Iterat...

hi tom, "disp "sigsqeta" %variance*exp(theta)" not disp "sigsqeta" %variance*theta Similarly for the others. do you mean that when i reduced the dimension of the numerical search by concentration of one of the variance, i used, sigmaeta=thetasigmaepsilon, in the model w...

Hello Mr Tom, Thank you very much for all your invalauble comments so far.. After so many trials, i discovered that the log form estimation may not be appropriate because the variances are quite small.. I decided to concentrate the measurement error out of the estimation.. This two seems to be the b...

Hi Tom, I am really indebted to you for all your help.. May God Almighty reward you.. Meanwhile, i effected the wrong ordering in Sw as detected by you.. The result obtained still suffers from the numerical problem.. i.e Nonlin lsigsqeps lsigsqeta lsigsqpsi lsigsqzeta stat(noprint) imp compute lsigs...

Hi Tom, Thank you very much for all the troubles.. meanwhile, i have tried estimating the model variances in log form but still the estimates still suffer from numerical problem..what would you recommend for these two output? linreg(noprint) imp # constant price implag compute sigsqeps=%seesq linreg...

Hi Tom, Thank you very much for your help.. pls could you check these two output for me.. what is your opinion about the estmates? (1) DLM - Estimation by BFGS Convergence in 3 Iterations. Final criterion was 0.0000000 <= 0.0000100 Quarterly Data From 1970:01 To 2009:03 Usable Observations 159 Rank ...

Hi Tom, Thank you very much for your suggestion.. pls, i have tried estimating the variances in log form but i noticed this error information.. could u please help me out? Nonlin sigsqeps sigsqeta sigsqpsi sigsqzeta stat(noprint) imp compute sigsqeps =log(.001*%variance),sigsqeta = log(5.1012e-006) ...

Hi, I am estimating a TVP model of the form yt = B0t + B1xi1+B2t xi2+et B0t = B0,t-1+ nt B1t = B1,t-1+ psit B2t = B2,t-1 + zetat The estimation of the hyper parameters yields: DLM - Estimation by BFGS NO CONVERGENCE IN 39 ITERATIONS LAST CRITERION WAS 0.0000097 ESTIMATION POSSIBLY HAS STALLED OR MAC...