Forum: help
Monitor Forum | 
Start New Thread
| RE: Efficiency Effects Frontier (BC1995) - multicollinearity issue [ Reply ] By: Maciej Wawrzyniak on 2022-06-10 18:09 | [forum:49410] |
|
Ok, thank you for your response. Kind Regards, Maciej, |
|
| RE: Efficiency Effects Frontier (BC1995) - multicollinearity issue [ Reply ] By: Arne Henningsen on 2022-06-06 04:48 | [forum:49404] |
| In many empirical applications, different combinations of the values for the intercept of the frontier model, the intercept of the inefficiency model, the coefficients of the Z variables, gamma, and sigma give almost the same log-likelihood value. Therefore, the maximisation of the log-likelihood value is difficult and the values of the above-mentioned coefficients that give the highest log-likelihood value depend on ‘chance’, which sometimes gives ‘crazy’ values. Removing the intercept of the inefficiency model often avoids this problem but forces the regression line of the inefficiency model to go through the origin, which could be inappropriate. | |
| Efficiency Effects Frontier (BC1995) - multicollinearity issue [ Reply ] By: Maciej Wawrzyniak on 2022-06-05 17:59 | [forum:49403] |
|
Dear Professor Henningsen, I would like for your assistance with one problem with Battese Coelli 1995 model specification (production function). When I try to estimate the model, my results indicates significant multicollinearity between variables which defines mi for the trunced normal distribution (Z-variables). I have noted that the issue appears only if intercept is one of the Z-variables. When I remove intercept, the issue is solved and I have 'correct' results. Nevertheless, I am trying to figure out why intercept influences entire estimation in such way. I have already tried to use different Z-variables, including randomly generated ones, however the issue is persistent regardles of the Z-variable. Efficiency Effects Frontier (see Battese & Coelli 1995) Inefficiency decreases the endogenous variable (as in a production function) The dependent variable is logged Iterative ML estimation terminated after 155 iterations: cannot find a parameter vector that results in a log-likelihood value larger than the log-likelihood value obtained in the previous step final maximum likelihood estimates Estimate Std. Error z value Pr(>|z|) (Intercept) 1.7802e-01 1.8690e-02 9.5247 < 2.2e-16 *** X1 9.3750e-01 3.3658e-02 27.8540 < 2.2e-16 *** X2 -5.3997e-02 3.0450e-02 -1.7733 0.0761714 . X3 1.7098e-01 3.6595e-02 4.6722 2.981e-06 *** I(0.5 * X1^2) 1.1515e-01 2.3160e-02 4.9718 6.632e-07 *** I(0.5 * X2^2) -1.4920e-01 3.8824e-02 -3.8430 0.0001215 *** I(0.5 * X3^2) -9.8769e-02 4.4204e-02 -2.2344 0.0254582 * I(X1 * X2) 1.7540e-02 2.2921e-02 0.7652 0.4441425 I(X1 * X3) -6.0276e-02 2.1489e-02 -2.8049 0.0050325 ** I(X2 * X3) 1.5337e-01 3.5876e-02 4.2749 1.912e-05 *** Z_(Intercept) -1.0326e+03 2.1415e+03 -0.4822 0.6296654 Z_Z1 2.4557e+02 5.0774e+02 0.4836 0.6286394 Z_Z2 -4.1985e+01 8.7169e+01 -0.4817 0.6300541 Z_Random 1.2408e+00 2.6088e+00 0.4756 0.6343451 sigmaSq 3.7261e+02 7.7141e+02 0.4830 0.6290816 gamma 9.9992e-01 1.7140e-04 5833.8684 < 2.2e-16 *** --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 log likelihood value: -141.772 cross-sectional data total number of observations = 368 mean efficiency: 0.7329988 Please find attached a sample from my dataset. Many thanks in advance, Kind Regards, Maciej, |
|
