Hi, I am trying to perform model 3: Half-normal stochastic production frontier model from the book "Stochastic frontier analysis using Stata" (page 292).
I would like to do the sfmodel command in order to do decomposition of TFP later. The first part is like this:
sfmodel lny, prod dist(h) frontier(lnl lnk lnl2 lnk2 lnllnk t tt tll tlk) usigmas(t) vsigmas()
sf_init, frontier(b1) usigmas(0 0) vsigmas(0)
ml max, difficult gtol(1e-5) nrtol(1e-5)
lnl is the log of labor, lnk log of capital, lnl2 and lnk2 is lnl*lnl*0.5 and similarly for lnk. t is a time variable (year in my case) and tt is t*t*0.5. Others are similar interactions.
The problem in the first step which becomes problem later is that t is omitted due to collinearity. I do not know why.
A second question is regarding usigmas(0 0) in the command sf_init. Then I get the answer: "initial vector: matrix must be dimension 12".
THanks in advance
I would like to do the sfmodel command in order to do decomposition of TFP later. The first part is like this:
sfmodel lny, prod dist(h) frontier(lnl lnk lnl2 lnk2 lnllnk t tt tll tlk) usigmas(t) vsigmas()
sf_init, frontier(b1) usigmas(0 0) vsigmas(0)
ml max, difficult gtol(1e-5) nrtol(1e-5)
lnl is the log of labor, lnk log of capital, lnl2 and lnk2 is lnl*lnl*0.5 and similarly for lnk. t is a time variable (year in my case) and tt is t*t*0.5. Others are similar interactions.
The problem in the first step which becomes problem later is that t is omitted due to collinearity. I do not know why.
A second question is regarding usigmas(0 0) in the command sf_init. Then I get the answer: "initial vector: matrix must be dimension 12".
THanks in advance
Comment