I am trying to perform a congestion analysis, using R (benchmarking package). I would like to calculate the TE under these 2 different assumptions. After going through dozens of papers, I am still confused as to how the TE ratio can be calculated. Looking at the Benchmarking package, I don't see any specific function for this purpose. Appreciate if any DEA experts can provide some insight.
Dear Foo,
I thought that it would be better to answer through your post rather than replying to your e-mail since other researches can also benefit from this and also they can probably add some more comments and suggestions.
The answer will be that instead of using Benchmarking you could use package FEAR available in the following link:
http://www.clemson.edu/economics/faculty/wilson/Software/FEAR/fear.html
You will need a licence which is free for academic purposes provided by professor Paul Wilson. You need to follow carefully the instructions.
Now concerning your question when calculating DEA efficiency estimates, the function dea gives you the option to define either strong or weak disposability through command DISP. Specifically, it is a character string equal to either "strong" (default) or "weak", indicating either strong disposability of inputs and outputs or weak disposability of inputs and outputs.
I hope that this helps
Best regards
Yiorgos
This is the manual which gives you instructions for the different functions in package FEAR
Really appreciate your kind reply Yiorgos !!
Please allow me to ask one more question here :
I am trying to compute the scale efficiency by dividing TE(CRS)/TE(VRS) - running 2 dea using CRS and VRS orientation. But i noticed that I cannot get R to indicate if it is IRS or DRS ( after knowing whether it is scale efficient or inefficient). Am I missing something ? Is that any way I can compute and find out if it is IRS or DRS ?
Thank you in advance !
Dear Foo,
When Scale Efficiency (SE)= CRS efficiency =VRS =1 efficiency then you have CRS
When VRS efficiency>SE efficiency>CRS and SE
Thank you so much Yiorgos!
Really appreciate your kind input and the material.
Dear Foo,
Something more about the DRS, IRS and CRS.
You can easily set a series of function to calculate that through the Benchmarking package.
First of all you need to solve DEA for 3 different programs
1. Efficiency under VRS
2. Efficiency under DRS (officially NIRS)
3. Efficiency under CRS
Then you need to calculate Scale Efficiency and use theorem 2 to decide about IRS, DRS or CRS.
Check this following code for R based on the Benchmarking package
Effvrs
Thanks Yiorgos, appreciate your clarification. I was in fact trying to built this algorithm using your explanation earlier. However, I guess bcos i am still pretty new to R and programming, thus yet to figure out after many attempts :-(. Thrilled to see this "answer" !
May I ask what does "1e-4"mean?
Another thing, I have tried to contact Paul W Wilson thru his email stated on the package, the author of the FEAR packages without success ; no reply after a week so far. What is your experience like?
Hello again!
1e-4=0.0001 i.e. the notation is used to symbolize small numbers. The number at the the end identifies the decimal points to 0.
About the second question I guess you will need to wait some time. Try to send him a reminder after a week.
I was lucky and I had a quick response.
Hi !
Sorry to keep bothering, just like to clarify one last piece for the above equation :
Is DRS-VRS
Hello.
I will generally set it to a number less than 1e-4 and not more than that.
And yes we try to approximate the true zero here.
Best regards
Yiorgos
Dear Foo,
So I had a thought over the previous code and I thought that some improvements can be done.
This is an updated version of the previous code which can also give straight information about IRS, DRS and CRS.
# Load required R packages
library(Benchmarking)
library(xlsx) # you might need to install this package if you haven't already
Effvrs
Hello again!
Yes definitely just I have build up as a logical test the full theorem 2 of Banker. You can omit se if you like. The result will not be affected.
Dear may be this link is useful for you http://www.uq.edu.au/economics/cepa/software.php
Thanks Carlos ! Came across that b4, but yet to have time to play around with it. Mind sharing your experience of using it?
Dear I sugesst you to use the software Stochastic Frontier 4.1 it is more easy to use for finding ET it may be organised in panel data. The other software is DEA 2.1 with Malmquist Index where you can to fin Produtivity change, ET chage, scale ET and pure ET change.
Carlos