"Jim Simmons" <emailjimsimmons@YAHOO.COM> ha scritto nel messaggio news:20050424180743.86963.qmail@web81607.mail.yahoo.com... > Grado, > Or would you also like to screen for any non-normal > distribution such as highly skewed ones, or ones with kurtosis > problems so extreme that they are non-normal? >

I just want to check for bimodal distribution of prices, because it may mean there are classification problem over an item by some retailer (i.e. retailer A codifies with EAN code xxxxx an item with average price 10, while retailer B codiefies with the same EAN an item with average price 1000).

> Also, when you say you are checking over a thousand items do you > mean you have to check a thousand variables(prices) in a single > data set, in multiple data sets, or a single variable in more > than a thousand data sets, or some combination of these > conditions?

No, I'm checking for a thousand of distinct items with no more than 50/60 prices. So it is an extensive check without so much obs.

>What type of automated precedure is currently in use > for this task? Macro's, SCL?

Just a simple macro. At the moment I am only highlighting extreme prices, the ones out 5th and 95th percentile, but this approach is not adequate for bimodal distributions.

Thank you for any suggestion.