Re: [Gretl-users] VAR - bug report

Thanks, Sorry, I have indeed sent identical scripts. Now, I attach two parts of scripts and the output of each of the scripts. The titles of each section below are Version A) Version A output) Version B) The difference between Version A) and Version B) is the location of the "0" in the definition of the system. In Version A) the "0" are the first independent variables and in version B) the "0" are the last independent variable. Comparing the output of the unrestricted VAR for both versions shows the difference in the coefficients and the "shift" in results. As far as I can tell, Version A results are correct and Version B results are shifted. I hope this helps. With kind regards, Ofer Cornfeld TAU Version A) system method = fiml equation p_d 0 p_d(-1) d_d(-1) r(-1) equation d_d 0 p_d(-1) d_d(-1) r(-1) equation r 0 p_d(-1) d_d(-1) r(-1) end system restrict --quiet 1 - rho*b[1,2] - b[2,2] + b[3,2] = 0 -b[2,3] + b[3,3] - rho*b[1,3] = 0 -b[2,4] + b[3,4] - rho*b[1,4] = 0 end restrict estimate method=fiml Version A Output) Endogenous variables: p_d d_d r Predetermined variables: p_d(-1) d_d(-1) r(-1) Exogenous variables: const Equation system, Full Information Maximum Likelihood Convergence achieved after 1 iterations Log-likelihood = 1101.98 Equation 1: FIML, using observations 1976:3-2007:3 (T = 125) Dependent variable: p_d coefficient std. error z p-value -------------------------------------------------------- const -0.0207539 0.0231735 -0.8956 0.3705 p_d_1 1.00370 0.00638197 157.3 0.0000 *** d_d_1 -0.198321 0.0625185 -3.172 0.0015 *** r_1 0.161816 0.104042 1.555 0.1199 Mean dependent var 3.517098 S.D. dependent var 0.194763 Sum squared resid 0.022080 S.E. of regression 0.013291 R-squared 0.995306 Adjusted R-squared 0.995189 Equation 2: FIML, using observations 1976:3-2007:3 (T = 125) Dependent variable: d_d coefficient std. error z p-value -------------------------------------------------------- const -0.0421234 0.0334717 -1.258 0.2082 p_d_1 0.0122965 0.00921807 1.334 0.1822 d_d_1 0.110828 0.0903013 1.227 0.2197 r_1 0.0149595 0.150278 0.09955 0.9207 Mean dependent var 0.001698 S.D. dependent var 0.019565 Sum squared resid 0.046065 S.E. of regression 0.019197 R-squared 0.029555 Adjusted R-squared 0.005494 Equation 3: FIML, using observations 1976:3-2007:3 (T = 125) Dependent variable: r coefficient std. error z p-value ------------------------------------------------------- const 0.0757555 0.0191680 3.952 7.74e-05 *** p_d_1 -0.0159735 0.00527884 -3.026 0.0025 *** d_d_1 -0.135023 0.0517121 -2.611 0.0090 *** r_1 0.121502 0.0860582 1.412 0.1580 Mean dependent var 0.021934 S.D. dependent var 0.012110 Sum squared resid 0.015107 S.E. of regression 0.010993 R-squared 0.169346 Adjusted R-squared 0.148751 Cross-equation VCV for residuals (correlations above the diagonal) 0.00017664 (-0.120) (0.651) -3.0525e-005 0.00036852 (-0.157) 9.5124e-005 -3.3213e-005 0.00012085 log determinant = -26.1454 Restriction set 1: -0.97137*b[1,2] - b[2,2] + b[3,2] = -1 2: -b[2,3] + b[3,3] - 0.97137*b[1,3] = 0 3: -b[2,4] + b[3,4] - 0.97137*b[1,4] = 0 Endogenous variables: p_d d_d r Predetermined variables: p_d(-1) d_d(-1) r(-1) Exogenous variables: const Equation system, Full Information Maximum Likelihood Convergence achieved after 1 iterations Log-likelihood = 1101.79 Equation 1: FIML, using observations 1976:3-2007:3 (T = 125) Dependent variable: p_d coefficient std. error z p-value -------------------------------------------------------- const -0.0195455 0.0228832 -0.8541 0.3930 p_d_1 1.00339 0.00630179 159.2 0.0000 *** d_d_1 -0.203430 0.0617331 -3.295 0.0010 *** r_1 0.156953 0.102735 1.528 0.1266 Mean dependent var 3.517098 S.D. dependent var 0.194763 Sum squared resid 0.022082 S.E. of regression 0.013291 R-squared 0.995305 Adjusted R-squared 0.995189 Equation 2: FIML, using observations 1976:3-2007:3 (T = 125) Dependent variable: d_d coefficient std. error z p-value ------------------------------------------------------- const -0.0323349 0.0155827 -2.075 0.0380 ** p_d_1 0.00978208 0.00427090 2.290 0.0220 ** d_d_1 0.0694417 0.0418382 1.660 0.0970 * r_1 -0.0244299 0.0696262 -0.3509 0.7257 Mean dependent var 0.001698 S.D. dependent var 0.019565 Sum squared resid 0.046180 S.E. of regression 0.019221 R-squared 0.028488 Adjusted R-squared 0.004401 Equation 3: FIML, using observations 1976:3-2007:3 (T = 125) Dependent variable: r coefficient std. error z p-value ------------------------------------------------------- const 0.0741333 0.0185286 4.001 6.31e-05 *** p_d_1 -0.0155568 0.00510228 -3.049 0.0023 *** d_d_1 -0.128164 0.0499825 -2.564 0.0103 ** r_1 0.128030 0.0831798 1.539 0.1238 Mean dependent var 0.021934 S.D. dependent var 0.012110 Sum squared resid 0.015110 S.E. of regression 0.010994 R-squared 0.169244 Adjusted R-squared 0.148647 Cross-equation VCV for residuals (correlations above the diagonal) 0.00017665 (-0.119) (0.651) -3.0411e-005 0.00036944 (-0.158) 9.5105e-005 -3.3367e-005 0.00012088 log determinant = -26.1422 LR test for the specified restrictions: Restricted log-likelihood = 1101.79 Unrestricted log-likelihood = 1101.98 Chi-square(3) = 0.399047 [0.9404] Version B) system method = fiml equation p_d p_d(-1) d_d(-1) r(-1) 0 equation d_d p_d(-1) d_d(-1) r(-1) 0 equation r p_d(-1) d_d(-1) r(-1) 0 end system restrict --quiet 1 - rho*b[1,2] - b[2,2] + b[3,2] = 0 -b[2,3] + b[3,3] - rho*b[1,3] = 0 -b[2,4] + b[3,4] - rho*b[1,4] = 0 end restrict estimate method=fiml Version B Output) Endogenous variables: p_d d_d r Predetermined variables: p_d(-1) d_d(-1) r(-1) Exogenous variables: const Equation system, Full Information Maximum Likelihood Convergence achieved after 3 iterations Log-likelihood = 1101.98 Equation 1: FIML, using observations 1976:3-2007:3 (T = 125) Dependent variable: p_d coefficient std. error z p-value -------------------------------------------------------- const 1.00370 0.00638197 157.3 0.0000 *** p_d_1 -0.198321 0.0625185 -3.172 0.0015 *** d_d_1 0.161816 0.104042 1.555 0.1199 r_1 -0.0207539 0.0231735 -0.8956 0.3705 Mean dependent var 3.517098 S.D. dependent var 0.194763 Sum squared resid 0.022080 S.E. of regression 0.013291 R-squared 0.995306 Adjusted R-squared 0.995189 Equation 2: FIML, using observations 1976:3-2007:3 (T = 125) Dependent variable: d_d coefficient std. error z p-value -------------------------------------------------------- const 0.0122965 0.00921807 1.334 0.1822 p_d_1 0.110828 0.0903013 1.227 0.2197 d_d_1 0.0149595 0.150278 0.09955 0.9207 r_1 -0.0421234 0.0334717 -1.258 0.2082 Mean dependent var 0.001698 S.D. dependent var 0.019565 Sum squared resid 0.046065 S.E. of regression 0.019197 R-squared 0.029555 Adjusted R-squared 0.005494 Equation 3: FIML, using observations 1976:3-2007:3 (T = 125) Dependent variable: r coefficient std. error z p-value ------------------------------------------------------- const -0.0159735 0.00527884 -3.026 0.0025 *** p_d_1 -0.135023 0.0517121 -2.611 0.0090 *** d_d_1 0.121502 0.0860582 1.412 0.1580 r_1 0.0757555 0.0191680 3.952 7.74e-05 *** Mean dependent var 0.021934 S.D. dependent var 0.012110 Sum squared resid 0.015107 S.E. of regression 0.010993 R-squared 0.169346 Adjusted R-squared 0.148751 Cross-equation VCV for residuals (correlations above the diagonal) 0.00017664 (-0.120) (0.651) -3.0525e-005 0.00036852 (-0.157) 9.5124e-005 -3.3213e-005 0.00012085 log determinant = -26.1454 Restriction set 1: -0.97137*b[1,2] - b[2,2] + b[3,2] = -1 2: -b[2,3] + b[3,3] - 0.97137*b[1,3] = 0 3: -b[2,4] + b[3,4] - 0.97137*b[1,4] = 0 Endogenous variables: p_d d_d r Predetermined variables: p_d(-1) d_d(-1) r(-1) Exogenous variables: const Equation system, Full Information Maximum Likelihood Convergence achieved after 250 iterations Log-likelihood = 345.006 Equation 1: FIML, using observations 1976:3-2007:3 (T = 125) Dependent variable: p_d coefficient std. error z p-value --------------------------------------------------------- const -0.0193537 0.0878860 -0.2202 0.8257 p_d_1 0.998764 0.109989 9.081 1.08e-019 *** d_d_1 -0.203183 0.183042 -1.110 0.2670 r_1 0.156279 0.0407694 3.833 0.0001 *** Mean dependent var 3.517098 S.D. dependent var 0.194763 Sum squared resid 1476.896 S.E. of regression 3.437320 R-squared 0.995305 Adjusted R-squared 0.995189 Equation 2: FIML, using observations 1976:3-2007:3 (T = 125) Dependent variable: d_d coefficient std. error z p-value ------------------------------------------------------- const -0.0323934 0.00966810 -3.351 0.0008 *** p_d_1 0.0111920 0.0882042 0.1269 0.8990 d_d_1 0.0693663 0.146788 0.4726 0.6365 r_1 -0.0242243 0.0326944 -0.7409 0.4587 Mean dependent var 0.001698 S.D. dependent var 0.019565 Sum squared resid 2.453318 S.E. of regression 0.140095 R-squared 0.028488 Adjusted R-squared 0.004401 Equation 3: FIML, using observations 1976:3-2007:3 (T = 125) Dependent variable: r coefficient std. error z p-value --------------------------------------------------------- const 0.0742611 0.0109248 6.797 1.06e-011 *** p_d_1 -0.0186386 0.0569082 -0.3275 0.7433 d_d_1 -0.127999 0.0947054 -1.352 0.1765 r_1 0.127581 0.0210940 6.048 1.46e-09 *** Mean dependent var 0.021934 S.D. dependent var 0.012110 Sum squared resid 16.65223 S.E. of regression 0.364990 R-squared 0.169244 Adjusted R-squared 0.148647 Cross-equation VCV for residuals (correlations above the diagonal) 11.815 (0.990) (-0.999) 0.47684 0.019627 (-0.991) -1.2537 -0.050654 0.13322 log determinant = -14.0337 LR test for the specified restrictions: Restricted log-likelihood = 345.006 Unrestricted log-likelihood = 1101.98 Chi-square(3) = 1513.96 [0.0000]