Forum: help
Monitor Forum | Start New ThreadRE: Constant efficiency if timeEffect=FALSE [ Reply ] By: Arne Henningsen on 2019-10-14 06:08 | [forum:47010] |
It sounds like your results are exactly as expected given the model specifications that you chose. I suggest that you take a look at different specifications of the SFA models, e.g., at [1] and/or at section 6 of [2]. [1] https://absalon.ku.dk/eportfolios/97/entries/124/files/eRLBmNpiTgduh7KxfYdqVZoUQkQNmddw315zb386 [2] https://leanpub.com/ProdEconR |
Constant efficiency if timeEffect=FALSE [ Reply ] By: Maciej Wawrzyniak on 2019-07-15 19:51 | [forum:46860] |
Dear All, I am workings with the frontier package for my PhD thesis and I have strange issue with the trans log function estimated using frontierQuad function. If i set timeEffect=FALSE I get an efficiency estimations constant for all T i.e. each observations have that same level of efficiency for each T. As far as I understand package and its functionality this option should lead to „unique” efficiency estimations for each combination observation-time, with no time effect. Initially, I thought that it can be some issue with my data, however if i set timeEffect=TRUE, i have a reasonable output. I would be very grateful, if somebody could assist me with this matter. I have information about 24 companies for 5 year (so n=24 and t=5). Number of actual observations is slightly smaller, as panel is unbalanced (In total i have 4 gaps In observations). Please find below some additionaly details regarding my data set structure. Observations T Output Input1 Input2 Input3 1 1 Log(Y) Log(x1) Log(x2) Log(x3) 1 2 ….. ….. ….. ….. 1 3 ….. ….. ….. ….. 1 4 ….. ….. ….. ….. 1 5 ….. ….. ….. ….. 2 1 ….. ….. ….. ….. 2 2 ….. ….. ….. ….. ….. ….. ….. ….. ….. ….. |