Most experimental research has assumed a uniform demand distribution and proved that the average orders differ significantly from the optimum. Let's assume two demand distributions - one a normal distribution with mean as 100 and standard deviation as 25, and other a uniform distribution varying between 25 and 175 units. Both distributions have a mean of 100 units. If we assume a 3 sigma variation for the normal distribution, both distributions would have a similar range.
A critical fractile of say 0.75 would mean an order of 117 units (difference of 17 from mean) in case of the normal distribution and it would mean an order of 138 units (difference of 38 from mean). Assuming participants anchor close to the mean, it would be easier in the case of Uniform distribution to prove significant difference than in case of normal distribution.
The uniform distribution should converge to the normal distribution in large sample replications. If you are thinking that they do not converge, then the demand phenomenon must be distributed according to another distribution--Poisson or Beta for instance. This can be verified through simulation.
I used the uniform distribution for housing demand, and found that they converge for single family and not rentals. The uniform was driving stable parameters such as household size, participation rates, institutional population, and tenure. However, since the uniform was approximately close to the normal, I accepted the estimates. I attach a file with that study.
A paper entitled " Bounded Rationality in Newsvendor Models" has shed light into this topic I suppose.
Take a look at it. You might find it useful.
- Badges
- Science topic
- Similar topics
- Chemistry
- Pharmacology
- Pharmaceuticals
- Drugs