02-16-2016 12:22 PM
This time I intend to use a VAR model to see how daily retail sales inteact with related website visits. Specifically, consumers visit the product website and then shop in retail stores. However, retail sales are always zeros because retail stores are closed on Sundays and holidays.
Any suggestions on how to handle this kinds of data in a VAR model?
Thanks as always!
02-16-2016 12:43 PM
I'm assuming this is an extension of this problem:
Is it the holidays that are now causing the issues?
02-16-2016 12:56 PM
Yes, it is an extension of the problem. But the problem now is to establish association between retail sales and website visits through VAR, not simply to make forecast for retail sales. Thus, even though there is no sales on Sundays or holidays, dropping Sundays or holidays is not going to solve the problem as website visits on Sundays and holidays matter -- they actually matter a lot.
Any suggestion will be greatly appreciated!
02-16-2016 04:38 PM
if 0 has to stay in the data, you will have to create a dummy variable to control the level change. Basically treating Sundays as an "event". You can then use VAR models to look at the correlation between the sales and web traffic.