I presently have data I'm trying to quantitatively describe as much as possible to reduce the "engineering judgment" required to make a decision.

Specifically, I'm taking unit prices of contract items that are bid and displaying them on a histogram. The goal is to determine a "average" unit price for an item, but several issues complicate the data.

The unit price variable is influenced by:

The year in which it was bid (typically increse each year)

The location where the contract is executed (varies)

The quantity of the contract item (inv. prop. to quantity amount)

Presently, I have loaded the unit price data into an Access database, which makes it possible to filter the data by year, location, and quantity, to help reduce the effects of these variables. The Access histogram typically shows a positively skewed distribution.

At present, I have in mind to analyze the data in these ways:

Calculate the interquartile mean

Calculate the interquartile standard deviation

Calculate the ratio of the deviation to the mean as a percentage (sigma / mu) * 100.

The goal is to be able to quantitatively answer the questions, "What's the average unit price" and "How 'stable' is the data" (how much variation is there?).

On the latter question, I expect to use the ratio of the standard deviation to the mean to issue a "Excellent - Good - Fair - Poor" judgement.

However, one other issue remains - that of multiple "clusters" in the histogram. That is, there may be a few "high points" throughout (perhaps because the data hasn't been adequately filtered), and it would be nice to be able to identify that. I recall that the coefficient of dispersion does that (stand. dev. ^2 / mean), but I think it requires the assumption of a Poisson distribution for the data, which I don't think is true here.

Anyway, for what I am seeking to accomplish, is this the right track? Is there anything I'm missing or trying to do that I shouldn't be?

Any comments or thoughts would be appreciated. I can provide some real data if it would help.

Thanks.

Joel