附件

没人回答

怎么有人下载数据，没人回答，追加论坛币到200

就Iterative Maximization Procedure而言
这个结果应该可以接受吧
model1: llf= 0.18548197E+04
model2: llf= 0.18565698E+04
两者也很接近
我以你的数据,在R执行的结果
供你参考
########MODEL1
lx1=read.csv("1.csv")
lx1Front <- sfa( logY~logK+logL+t.1+X.logK..2+X.logL..2+t.2+logKlogL+tlogK+tlogL |
                  z1+z2+z3+z4+z5, data = lx1 ,restartFactor = 0.9954 )                                                           
summary(lx1Front)   
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 93 iterations:
log likelihood values and parameters of two successive iterations
are within the tolerance limit
Multiplied the initial values 3 time(s) by 0.9954 before the search procedure could start

final maximum likelihood estimates
                    Estimate     Std. Error        z value   Pr(>|z|)   
(Intercept)    8.3264632e-01  3.4042700e-03  2.4458880e+02 < 2.22e-16 ***
logK          -7.4681174e-04  9.6890060e-04 -7.7078000e-01 0.44083579   
logL           2.5081914e-03  7.9909115e-04  3.1388100e+00 0.00169638 **
t.1            1.1511280e-03  2.5021425e-04  4.6005700e+00 4.2134e-06 ***
X.logK..2      2.9928358e-01  3.2985565e-04  9.0731684e+02 < 2.22e-16 ***
X.logL..2      2.4384756e-05  7.6299523e-05  3.1959000e-01 0.74927725   
t.2            7.8924228e-05  2.2967696e-05  3.4363100e+00 0.00058969 ***
logKlogL       1.3718646e-04  1.3009380e-04  1.0545200e+00 0.29164510   
tlogK          1.4645107e-05  4.6840712e-05  3.1266000e-01 0.75454078   
tlogL         -7.8084825e-04  6.5910796e-05 -1.1847050e+01 < 2.22e-16 ***
Z_(Intercept) -4.8173216e+00  3.5453410e-01 -1.3587750e+01 < 2.22e-16 ***
Z_z1           4.4747523e+00  2.9119260e-01  1.5366990e+01 < 2.22e-16 ***
Z_z2           1.0925479e+01  6.9746806e-01  1.5664490e+01 < 2.22e-16 ***
Z_z3          -2.4971168e+00  1.7257560e-01 -1.4469700e+01 < 2.22e-16 ***
Z_z4           1.4907022e-01  1.1232232e-02  1.3271650e+01 < 2.22e-16 ***
Z_z5           3.8517333e-02  4.7408876e-03  8.1245000e+00 4.4922e-16 ***
sigmaSq        1.4022896e+00  8.8060512e-02  1.5924160e+01 < 2.22e-16 ***
gamma          9.9999999e-01  7.6857965e-09  1.3011013e+08 < 2.22e-16 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
log likelihood value: 1896.21601

panel data
number of cross-sections = 287
number of time periods = 6
total number of observations = 1722
thus there are 0 observations not in the panel

mean efficiency of each year
           1            2            3            4            5            6
0.8681691083 0.8908109866 0.8986839255 0.9033270312 0.9047456110 0.8995385629

mean efficiency: 0.8942125376

#########MODEL2
lx1=read.csv("1.csv")
lx2Front <- sfa( logY~logK+logL+t.1+X.logK..2+X.logL..2+t.2+logKlogL+tlogK+tlogL |
                  z5+z3+z1+z2+z4, data = lx1 ,restartFactor = 0.9954 )                                                        
summary(lx2Front)  
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 95 iterations:
log likelihood values and parameters of two successive iterations
are within the tolerance limit
Multiplied the initial values 3 time(s) by 0.9954 before the search procedure could start

final maximum likelihood estimates
                    Estimate     Std. Error        z value   Pr(>|z|)   
(Intercept)    8.3031083e-01  3.4693193e-03  2.3932961e+02 < 2.22e-16 ***
logK          -1.7653763e-03  5.3324914e-04 -3.3106000e+00 0.00093095 ***
logL           6.9835171e-03  1.1181974e-03  6.2453300e+00 4.2289e-10 ***
t.1           -7.1591147e-04  4.0330342e-04 -1.7751200e+00 0.07587828 .  
X.logK..2      3.0060272e-01  7.4176716e-04  4.0525213e+02 < 2.22e-16 ***
X.logL..2      1.7661509e-04  3.9555749e-05  4.4649700e+00 8.0081e-06 ***
t.2            6.6641980e-05  2.0438577e-05  3.2606000e+00 0.00111178 **
logKlogL      -6.3611411e-04  1.4860125e-04 -4.2806800e+00 1.8632e-05 ***
tlogK          2.2242445e-04  5.5921288e-05  3.9774600e+00 6.9657e-05 ***
tlogL         -5.5472693e-04  7.2066538e-05 -7.6974300e+00 1.3883e-14 ***
Z_(Intercept) -4.0420022e+00  3.2241049e-01 -1.2536820e+01 < 2.22e-16 ***
Z_z5           3.7624590e-02  5.5473322e-03  6.7824700e+00 1.1814e-11 ***
Z_z3          -2.3457801e+00  2.2516509e-01 -1.0418050e+01 < 2.22e-16 ***
Z_z1           4.0332752e+00  3.3640429e-01  1.1989370e+01 < 2.22e-16 ***
Z_z2           9.7118836e+00  7.5336285e-01  1.2891380e+01 < 2.22e-16 ***
Z_z4           1.4178236e-01  1.5851049e-02  8.9446700e+00 < 2.22e-16 ***
sigmaSq        1.2514380e+00  9.7753805e-02  1.2801940e+01 < 2.22e-16 ***
gamma          9.9999999e-01  6.6013139e-09  1.5148499e+08 < 2.22e-16 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
log likelihood value: 1897.780625
panel data
number of cross-sections = 287
number of time periods = 6
total number of observations = 1722
thus there are 0 observations not in the panel

mean efficiency of each year
           1            2            3            4            5            6
0.8690837444 0.8914485522 0.8989932306 0.9032905001 0.9043624319 0.8988700341

mean efficiency: 0.8943414155

虽然epon没有完全回答我的问题，但很感谢用r给出结果，从r结果看，这也从另一个角度证实z的排列顺序对SFA模型结果确实有一定影响