Hi,
I have been estimating several SFA production functions (using spaniel), mostly with success (I think) but have run into an issue when I attempted to estimate two using the Kumbhakar (1990) model option. The data I have is a balanced panel dataset across 9 years with 21 hospitals.
The model has 3 inputs to production. When I estimate the model using a Cobb-Douglas functional form, everything seems to be ok, however when I estimate a Translog function I seem to run into a problems. As you can see below there are some missing values in the regression output for "loglabor2" and "_cons" - I'm wondering if this is being caused by issues related to multicollinearity of my independent variables (they are reasonably highly correlated). Has anyone experienced this before? Any suggestions on what might be going on and what I could do about it?
I have been estimating several SFA production functions (using spaniel), mostly with success (I think) but have run into an issue when I attempted to estimate two using the Kumbhakar (1990) model option. The data I have is a balanced panel dataset across 9 years with 21 hospitals.
The model has 3 inputs to production. When I estimate the model using a Cobb-Douglas functional form, everything seems to be ok, however when I estimate a Translog function I seem to run into a problems. As you can see below there are some missing values in the regression output for "loglabor2" and "_cons" - I'm wondering if this is being caused by issues related to multicollinearity of my independent variables (they are reasonably highly correlated). Has anyone experienced this before? Any suggestions on what might be going on and what I could do about it?
Code:
. sfpanel logWIES loglabor logbeds logtotclinical, model(kumb90);
initial: Log likelihood = -150.6797
Iteration 0: Log likelihood = -150.6797 (not concave)
Iteration 1: Log likelihood = 255.61181
...
Iteration 11: Log likelihood = 281.17544
Iteration 12: Log likelihood = 281.17544
Time-varying parametric model (half-normal) Number of obs = 189
Group variable: DHB_new Number of groups = 21
Time variable: year Obs per group: min = 9
avg = 9.0
max = 9
Prob > chi2 = 0.0000
Log likelihood = 281.1754 Wald chi2(3) = 2599.10
--------------------------------------------------------------------------------
logWIES | Coef. Std. Err. z P>|z| [95% Conf. Interval]
---------------+----------------------------------------------------------------
Frontier |
loglabor | .1676521 .068641 2.44 0.015 .0331183 .302186
logbeds | .5180803 .0457991 11.31 0.000 .4283158 .6078448
logtotclinical | .269625 .0394514 6.83 0.000 .1923016 .3469484
_cons | 3.375925 .150471 22.44 0.000 3.081007 3.670843
---------------+----------------------------------------------------------------
Bt |
b | -2.770481 .8797331 -3.15 0.002 -4.494727 -1.046236
c | .2944086 .0978066 3.01 0.003 .1027112 .4861061
---------------+----------------------------------------------------------------
/sigmau_2 | .0427497 .0155593 2.75 0.006 .012254 .0732455
/sigmav_2 | .0019232 .0002125 9.05 0.000 .0015066 .0023398
-------------+----------------------------------------------------------------
sigma_u | .2067601 .0376265 5.50 0.000 .1106978 .270639
sigma_v | .0438545 .0024233 18.10 0.000 .0388154 .0483714
lambda | 4.714687 .0375086 125.70 0.000 4.641171 4.788202
------------------------------------------------------------------------------
. sfpanel logWIES loglabor logbeds logtotclinical loglabor2 logbeds2 logtotclinical2 loglaborbeds loglaborclinical logbedsclinical, model(kumb90);
initial: Log likelihood = -421.02705
Iteration 0: Log likelihood = -421.02705 (not concave)
Iteration 1: Log likelihood = 194.65576 (not concave)
Iteration 2: Log likelihood = 244.05984 (not concave)
...
Iteration 98: Log likelihood = 286.44102 (not concave)
Iteration 99: Log likelihood = 286.44176 (not concave)
Iteration 100: Log likelihood = 286.44248 (not concave)
Time-varying parametric model (half-normal) Number of obs = 189
Group variable: DHB_new Number of groups = 21
Time variable: year Obs per group: min = 9
avg = 9.0
max = 9
Prob > chi2 = 0.0000
Log likelihood = 286.4425 Wald chi2(7) = 4.42e+09
----------------------------------------------------------------------------------
logWIES | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-----------------+----------------------------------------------------------------
Frontier |
loglabor | -1.844341 . . . . .
logbeds | 2.440416 .3303281 7.39 0.000 1.792985 3.087847
logtotclinical | .5751731 .1930512 2.98 0.003 .1967997 .9535464
loglabor2 | -2.530396 . . . . .
logbeds2 | -.2403704 .0544843 -4.41 0.000 -.3471578 -.1335831
logtotclinical2 | -.5094808 .0597611 -8.53 0.000 -.6266105 -.3923511
loglaborbeds | 1.208227 .045438 26.59 0.000 1.11917 1.297284
loglaborclinical | 1.362819 .0292809 46.54 0.000 1.305429 1.420208
logbedsclinical | -.9244875 .0393694 -23.48 0.000 -1.00165 -.8473249
_cons | 3.778695 . . . . .
-----------------+----------------------------------------------------------------
Bt |
b | -.2010611 .0984202 -2.04 0.041 -.3939611 -.0081612
c | .0317098 .01008 3.15 0.002 .0119533 .0514664
-----------------+----------------------------------------------------------------
/sigmau_2 | .0941397 .0366863 2.57 0.010 .0222359 .1660436
/sigmav_2 | .0018919 .000214 8.84 0.000 .0014725 .0023112
-------------+----------------------------------------------------------------
sigma_u | .306822 .0597844 5.13 0.000 .149117 .4074845
sigma_v | .0434958 .0024595 17.68 0.000 .0383735 .0480753
lambda | 7.054067 .0596304 118.30 0.000 6.937194 7.170941
------------------------------------------------------------------------------
