rizal yudha tama
2012-May-03 08:29 UTC
[R] MLE for estimating the parameters of the TVECM in R
Dear Mr. Matthieu Stigler i so excited for your package 'tsDyn'. firstly introduce myself, i student at Gadjah Mada University,Indonesia. i'am new user of R and applying it for solving Bi-Variate ( interest rate and inflation ) with threshold vector error correction model. now, i writing my final examination about threshold vector error correction model and i use refference from paper "testing for two regime threshold cointegration in vector error correction model" by hansen and seo (2002) to estimate parameter. i have tried to reduce MLE , and it's succes. now i have A1(hat), A2(hat) with MLE and gamma(hat), beta(hat) with grid search from MLE. My problem, when i using function HanSeo_TVECM() in R, this function can't running, only to estimate linier cointegration (VECM). and if i using packade tsDyn version 0-8.1, function HanSeo_TVECM() not avaliable. however there are function TVECM() but this function using CLS for estimate parameter. whether the MLE and CLS estimation would result in same relative values? can u help me sir? for function HanSen_TVECM()? thanks a lot output R from function HanSeo_TVECM> HanSeo_TVECM(data,lag=2,trim0.05,gn=300,bn=300)############################ ###Linear VECM estimates (by Johansen MLE) ############################ Cointegrating vector 1 -8.434287 Standard error of the cointegrating value 2.616888 Parameters ECM Intercept bi -1 inf -1 bi -2 Equation bi -0.008627598 -0.006055985 0.758366 0.02512693 -0.1240975 Equation inf 0.131374112 -0.355994435 3.971587 0.20726565 -2.7661240 inf -2 Equation bi -0.007603223 Equation inf -0.224955971 Standard errors with Eicker-White covariance matrix estimator ECM Intercept bi -1 inf -1 bi -2 inf -2 Equation bi 0.004081985 0.01663758 0.08914615 0.02456267 0.08221155 0.01799844 Equation inf 0.034760850 0.08915086 1.66241171 0.19362012 0.93278372 0.06298056 Negative LL -183.1256 AIC -159.1256 BIC -160.4877Error in solve.default(t(zzj) %*% zzj) : system is computationally singular: reciprocal condition number 1.15007e-020 [[alternative HTML version deleted]]