Exogenous Features and Covariates about bsts-causal_impact HOT 4 OPEN

Hi bigfivecode,

Would the flag be intended as a dummy variable i.e. to investigate structural change, or simply to indicate a post-intervention period?

from bsts-causal_impact.

I would like to take into account that there was a shortage of stock(items in inventory) at some point in the pre-period and I was wondering if I could introduce a dummy variable for the dates when the shortage happened to reflect this event.
Or would you recommend a better solution?

Many thanks in advance

from bsts-causal_impact.

You could certainly add this as a discrete variable to represent a binary Y/N shortage, however, the model would simply estimate it's effect as a point estimate i.e. it has no gauge of the relative level of inventory. IMHO, I think adding the inventory as a covariate in its continous form would be most effective. If you are simply trying to estimate the effect that an inventory shortage has on a dependent variable through time, then you might consider a structural time series as an alternative?

from bsts-causal_impact.

Thanks for the clarification. It makes perfect sense.

from bsts-causal_impact.