Proper way to analyse purchase data I have a dataset containing information about various factors influencing the amount purchased of different products.
factors include the number of people who bought it, the degree to which the product was on sale (90 or 80%), the number of times the product was on sale, etc.
I would like to see how much such factors influence the popularity of each product (number of people buying it)
I was wondering if a simple regression or anova would do the job or whether there are some other specific statistics more suitable for such problems.
further problems: I have data from different companies offering different products. I want to make inferences about the relationship between the type of product and the amount purchased by pooling all the data available from the different companies. However, my problem is that the different companies differ in size and therefore the amount they have sold. Can this be solved somehow?
 A: You would have to do tests for heteroscedasticity and multicollinearity. Then you could run a multitude of regression models which you would read up on to see which fits best, ARCH, OLS, GARCH...And if IIRC, you either take the log of the dependent variable or the log of all the variables so that when you take your regression, the coefficients give you the percent effect (influence) of each independent variable.
A: Regression sounds like a good starting place. You could also look at non-parametric methods like bootstrapping. 
It sounds like you might also be interested in data mining techniques if you're working with this kind of data. For example, if you have information about which products were bought together by the same customer, you can use the "apriori association mining" algorithm to find patterns. This method is often used to select the placement of pairs of items in stores, with a classic example being to put the diapers next to the beer (when new fathers are sent shopping alone, turns out they like to get a treat for themselves). 
