Dear Sam,

Here are a few starting pointers to your question.

· Units (production or decision making units (DMUs)) on the data envelope have 100 percent efficiency. They create a benchmark for other DMUs. As a start you can tabulate the score you get in a table rank say by the highest to the lowest.

· Units below the frontier are less than 100% efficient. They are assigned an efficiency score depending on how far they are below the envelope. DEA uses a sort of waited average of other DMU on the envelope to score units below them.

· The efficiency scores are calculated by the use of an optimizing technique, which can be programmed with a Linear Programming software. Solver in Excel can be programmed as such. But a software such as Frontiers and other on the internet can do more complex problems. For instance, you can use many different assumptions about returns to scale—constant, variable, non-increasing, non-decreasing and generalized. It appears that you have already done some programming on your data.

· An elementary comparison would require some sort of monotonic comparison where a higher score is preferred to a lower one. Calculating the scores is a complex job better left to softwares.

dear sir,

if I understand your question correctly, you would like to choose some efficient DMUs from a set of efficient DMUs. for this aim, you can use super-efficiency data envelopment analysis (super DEA). this model gives you the possibility of ranking efficient DMUs (not all DMUs).

I wish it could be helpful...

Bests.

Ehsan

You will need to look at their shadow prices or multipliers, as even though the combination scores of the two efficient DMUs are equal (and equal to unity), the way they combine the inputs and outputs are different - there you can discuss about their strengths and weaknesses.

Yes Ehsan is right. Compute super efficiencies and you will get a ranking of efficient units as well. You can also compute cross efficiencies to compare them. Returns to scale will also give some insight into their performance. Also compare for how many inefficient units are these units the peers. This also gives a ranking structure to the efficient units.