frontier4.1在输入数据时不手动取对数，软件在计算时会自动取对数么（在ins文件里选n

各位大神，我想请问就是同一套数据，我输入原数据，在ins中LOGGED DEPENDENT VARIABLE (Y/N)选择n，和我手动ln之后，选y出来的结果完全不一样呢，这是为啥呀
结果差别好大，该怎么调试呀
或者说以哪一个结果为主呀

这是取对数后的结果： Error Components Frontier (see B&C 1992)
The model is a production function
The dependent variable is logged


the ols estimates are :

                 coefficient     standard-error    t-ratio

  beta 0         0.30936882E+01  0.34323066E+00  0.90134377E+01
  beta 1         0.64501474E-02  0.27730026E-01  0.23260517E+00
  beta 2         0.65863985E-01  0.51742579E-01  0.12729166E+01
  beta 3         0.74920760E+00  0.30256347E-01  0.24761998E+02
  sigma-squared  0.39578661E+00

log likelihood function =  -0.30948013E+03

the estimates after the grid search were :

  beta 0         0.36470848E+01
  beta 1         0.64501474E-02
  beta 2         0.65863985E-01
  beta 3         0.74920760E+00
  sigma-squared  0.69717816E+00
  gamma          0.69000000E+00
   mu is restricted to be zero
   eta is restricted to be zero


iteration =     0  func evals =     20  llf = -0.30339361E+03
     0.36470848E+01 0.64501474E-02 0.65863985E-01 0.74920760E+00 0.69717816E+00
     0.69000000E+00
gradient step
iteration =     5  func evals =     44  llf = -0.30304547E+03
     0.34720646E+01-0.76109274E-02 0.75982777E-01 0.76615358E+00 0.70880087E+00
     0.71353443E+00
search failed. loc of min limited by rounding
iteration =     9  func evals =     92  llf = -0.30301024E+03
     0.34304773E+01-0.73617006E-02 0.73254217E-01 0.76840456E+00 0.68803626E+00
     0.69636921E+00


the final mle estimates are :

                 coefficient     standard-error    t-ratio

  beta 0         0.34304773E+01  0.32332497E+00  0.10609998E+02
  beta 1        -0.73617006E-02  0.25862943E-01 -0.28464280E+00
  beta 2         0.73254217E-01  0.49814341E-01  0.14705448E+01
  beta 3         0.76840456E+00  0.28376671E-01  0.27078742E+02
  sigma-squared  0.68803626E+00  0.89541684E-01  0.76839772E+01
  gamma          0.69636921E+00  0.76314764E-01  0.91249606E+01
   mu is restricted to be zero
   eta is restricted to be zero

log likelihood function =  -0.30301024E+03

LR test of the one-sided error =   0.12939790E+02
with number of restrictions = 1
[note that this statistic has a mixed chi-square distribution]

number of iterations =      9

(maximum number of iterations set at :   100)

number of cross-sections =    326

number of time periods =      1

total number of observations =    326

thus there are:      0  obsns not in the panel


以下是没有取对数，在ins里选n的结果：
Error Components Frontier (see B&C 1992)
The model is a production function
The dependent variable is not logged


the ols estimates are :

                 coefficient     standard-error    t-ratio

  beta 0        -0.48340960E+05  0.22056912E+06 -0.21916467E+00
  beta 1         0.94463413E+03  0.27249675E+03  0.34665887E+01
  beta 2         0.40992561E+05  0.24716961E+05  0.16584790E+01
  beta 3         0.50795307E+00  0.28266396E-01  0.17970210E+02
  sigma-squared  0.79023831E+13

log likelihood function =  -0.53013658E+04

the estimates after the grid search were :

  beta 0         0.45823860E+06
  beta 1         0.94463413E+03
  beta 2         0.40992561E+05
  beta 3         0.50795307E+00
  sigma-squared  0.80620442E+13
  gamma          0.50000000E-01
   mu is restricted to be zero
   eta is restricted to be zero


iteration =     0  func evals =     20  llf = -0.53016286E+04
     0.45823860E+06 0.94463413E+03 0.40992561E+05 0.50795307E+00 0.80620442E+13
     0.50000000E-01
gradient step
iteration =     5  func evals =     93  llf = -0.53016078E+04
     0.45823854E+06 0.92882359E+03 0.40992200E+05 0.50782773E+00 0.80620442E+13
     0.43096593E-01
search failed. loc of min limited by rounding
iteration =     6  func evals =     97  llf = -0.53016078E+04
     0.45823854E+06 0.92882359E+03 0.40992200E+05 0.50782773E+00 0.80620442E+13
     0.43096593E-01


the final mle estimates are :

                 coefficient     standard-error    t-ratio

  beta 0         0.45823854E+06  0.13148648E+01  0.34850620E+06
  beta 1         0.92882359E+03  0.22673456E+03  0.40965241E+01
  beta 2         0.40992200E+05  0.52766433E+01  0.77686131E+04
  beta 3         0.50782773E+00  0.17602053E-01  0.28850483E+02
  sigma-squared  0.80620442E+13  0.10000000E+01  0.80620442E+13
  gamma          0.43096593E-01  0.24647158E-01  0.17485421E+01
   mu is restricted to be zero
   eta is restricted to be zero

log likelihood function =  -0.53016078E+04

the likelihood value is less than that obtained
using ols! - try again using different starting values

number of iterations =      6

(maximum number of iterations set at :   100)

number of cross-sections =    326

number of time periods =      1

total number of observations =    326

thus there are:      0  obsns not in the panel

遇到了同样的问题，不知楼主是否已经解决？

同问，请问楼主解决了吗？

abc常启国 发表于 2018-6-8 16:05
遇到了同样的问题，不知楼主是否已经解决？

请问你后来这个问题解决了吗