0
$\begingroup$

What would be the best suited method to analyze the following:

 Week    Location     Revenue($MM)     TrainingEvent
 1       abc          1.2              0
 2       abc          1.2              1
 3       abc          1.6              0
 4       abc          1.5              0
 ....    abc          ...              ... 
 1       xyz          2.1              0
 2       xyz          2.0              0
 3       xyz          2.2              1
 4       xyz          2.9              0
 ...     xyz          ...              ...

The idea is that there are several hundred locations and 1 year worth of weeks, and training is conducted periodically to coach employees on proper sales techniques. I want to quantify the impact of training on revenue in future weeks. Rather than performing several hundred linear regressions with lagged variables and/or time series ARIMA type models, and somehow determining an average effect after standardizing the revenue variable, I'd like to think there is some sort of method that will essentially allow me to do all of this as one model.

Any suggestions?

$\endgroup$
  • $\begingroup$ Can you look at the sales of the trained employees, or do you only have aggregate data for each location? $\endgroup$ – gung Jul 10 '15 at 15:09
  • $\begingroup$ Yes, only the aggregate. I believe these are large group training exercises that take place in the mornings before the stores open. $\endgroup$ – Brandon Metzger Jul 10 '15 at 15:18
1
$\begingroup$

Set up 1 estimation where Y is related to the X (promo) variable . Estimate the effect globally essentially using pooled cross-sectional procedures. If there is a lagged effect of prior sales then care should be taken to incorporate such a structure. Software is available to do this including AUTOBOX which I helped write. Care should be taken to empirically identify and treat anomalous data rather than blithely ignoring them because you didn't know about then before you observed the data aka as the head in the sand approach.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.