[EMAIL PROTECTED] (jennl) wrote in message news:<[EMAIL PROTECTED]>... > I am first trying to build a model using regression analysis to > explain weekly sales, stripped of all exogenous variables (pricing, > seasonality, etc) and all marketing expenditures
You shouldn't strip or pre-filter these exogenous variables as their impact or weight is most likely inter-dependent with your three additional variables. Further more there may be lag effects of these stage 1 variables that you might not be correctly forming. Additionally there may be Interventions such as Level Shifts or Local time trends or Holiday Events which might change the Y variable before,on or after these events. In addition since you are dealing with time series there may be autoregressive structure in the errors reflecting omitted variables and one might need to add an ARIMA component and/or weekly dummies reflecting a deterministic effect for the week of the year. One of our customers Anheuser-Busch ( http://www.autobox.com/stories.html)is using our software (AUTOBOX) to perform a similar analysis . They are analyzing sales of 50 SKU's at 650,000 retail outlets (stores) by day incorporating price,weather,holidays,events abd special promotions. http://www.autobox.com/details.htm report http://www.autobox.com/actfore.jpg http://www.autobox.com/intrvent.html http://www.autobox.com/equation.txt http://www.autobox.com/verbal.txt http://www.autobox.com/stat.html Specifically they include a 0/1 variable for off/on promotion and price to customer in order to estimate the impact of promotion and the impact of price. If you wish to discuss this further please feel free to give me a call at 215-675-0652 Dave Reilly Automatic Forecasting Systems http://www.autobox.com Regards . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
