Continuous interventions?

[yangya...@gmail.com]

Can this methodology be adapted to handle continuous interventions rather than just a single period binary intervention variable? i.e. levels of advertising over time instead of the start of an advertising campaign?

[Kay Brodersen]

Hi yangyang2000,

You could construct a custom bsts model and add your metric of advertising intensity as a regressor. The posterior distribution over that regressor's coefficient will then provide you with an estimate of the effect size, provided that you have regressed out other sources of variation.

Kay

On 12 April 2015 at 02:51, <yangya...@gmail.com> wrote:

Can this methodology be adapted to handle continuous interventions rather than just a single period binary intervention variable? i.e. levels of advertising over time instead of the start of an advertising campaign?

--
You received this message because you are subscribed to the Google Groups "CausalImpact" group.
To unsubscribe from this group and stop receiving emails from it, send an email to causalimpact...@googlegroups.com.
To post to this group, send email to causal...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/causalimpact/b493059c-19a3-4eb8-961c-23a3d672dd0f%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.