I am using Data Envelopment Analysis (DEA) and Stochastic Frontier Analysis (SFA) to measure the eco-efficiency of dairy farms. I would like to use the quantity of chemical fertilizers as an input. Some farms, e.g. organic farms, do not apply any chemical fertilizers and have a 0 value for this variable. How can I deal with 0 values in DEA and SFA inputs and outputs? Is it correct if I replace the 0 with a 1 (knowing that 1 is a very small value for total fertilizer application on a farm and so it is very similar to having no fertilizer application at all)?
I think the problem with SFA, especially if you are using frontier, is missing values , otherwise you can use 0s.
Read this article
https://link.springer.com/chapter/10.1007/978-0-387-71607-7_17
You have two options:
-May use 1 instead of 0 if it is significantly smaller other collection values.
or
- Both models can handle 0 or missing data but you need to find appropriate codes for this situation.
In case of DEA, 0 is acceptable as output
In case of SFA, most of the papers added minimum value +1 in that input/output of each DMU (if you have even negative values), in case of only 0, I think +1 is better solution.
Please consider literature on it.
Thank you all for your explanations. The article mentioned by Anup Kumar Yadava was very useful and helped me find a solution to deal with zero values (i.e., using translation to eliminate zero or negative values, or using a small positive value).
Please check the following:
https://www.sciencedirect.com/topics/economics-econometrics-and-finance/data-envelopment-analysis
Article Alternative Methods to Examine Hospital Efficiency: Data Env...
https://scholar.cu.edu.eg/?q=basmaezzat/publications/developing-stochastic-input-oriented-data-envelopment-analysis-modelhttps://papers.ssrn.com/sol3/papers.cfm?abstract_id=870553
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size