"teradial" package - How to obtain reference weights/lambdas? (Data Envelopment Analysis)

[Arjan Trinks]

Dear all,

I want to run a standard DEA model to evaluate the technical efficiency (Theta) of a sample of firms. Specifically, this is an input-oriented BCC model (R.D. Banker, A. Charnes, W.W. Cooper, 1984. Some models for the estimation of technical and scale inefficiencies in Data Envelopment Analysis. Management Science, 30, 1078-1092)).

There are two commands that do this in Stata:
- dea (Ji and Lee, 2010: https://www.stata-journal.com/articl...article=st0193)
- teradial (Badunenko and Mozharovskyi, 2016: https://www.stata-journal.com/articl...article=st0444)

=> I wonder whether the command "teradial" is able to return the reference weights (lambdas) used to obtain the efficiency score.

A great benefit of "teradial" over "dea" is its speed. However, "dea" gives a matrix "r(dearslt)" that includes efficiency scores, references, slacks, and more...
Even though the authors of "teradial" explain that the code offers more capabilities, it does not seem give information on references and slacks...

Hopefully, somebody could clarify.

Arjan

[Phil Bromiley]

You didn't get a quick answer. You'll increase your chances of a useful answer by following the FAQ on asking questions.

With a user-written routine, unless someone happens to have used it extensively, you are unlikely to get the kind of detailed help you want. I assume you've looked over the teradial documentation very carefully. If so, you have two options. First, since teradial is an ado, you can look at the file to see if you can add something to return the references and slacks. Second, contact the author.

[Arjan Trinks]

Dear dr Phil,
Thanks. I indeed contacted the authors first; still no reaction, however.
I think the program hasn't been used much around here.
I didn't learn anything from going through the "teradial" documentation multiple times, and didn't succeed in making changes to the ado-file.

For now, I stick with the original "dea" package, which has a major drawback of it being terribly slow when the number of DMUs increases (for my analysis, it took 7 full days on my laptop).
Fortunately, I've got lucky by getting my hands on a supercomputer which runs the analysis in 1 hour.