Hi everybody
I'm wondering if we can use logged inputs and outputs to perform a DEA efficiency assessment. I think there will be no problem on the theoretical point of view since DEA is an optimization process and transforming the data up to a strict monotonic function (the logarithm here) doesn't affect the optimal solution. I'am asking because I have tried with a data set both specifications and the results are dramatically different: The efficiency scores obtained from the logged data set have higher average and lower dispersion.
Can anyone help?
What do you mean by logged input and output? It is the input that should be logged. The output may be transformed back to the original units. If so, the efficiency scores should not be different from the original case since the logarithm function is a monotonic increasing function.
Thank you for your answers. What youbhave suggested was important endeed to have more insight about the issue. Since the frontier inndea is a piecewise linear functions, when dealing with logged data the frontier will be a piecewise log-linear functions. This assume a cobb-douglass shape of the constraints in the linear programming and the 1 at the rhs must be transformed to be log(1) = 0 among others.
- Badges
- Science topic