I am trying to gain an understanding of the CausalImpact package and the BSTS approach. Consider the following model:

$y_t = \mu_t + z_t + \epsilon_t$, where $\epsilon_t \sim N(0,s^2)$, and $\mu_{t+1} = \mu_t + \eta_t$, where $\eta_t \sim N(0,s'^2)$, and $z_t = \beta x_t$ where $x_t$ is the control time series.

Am I correct in thinking that the default mode in the CausalImpact package employs the above model?

Now, with the default options, if I explicitly incorporate the time $t$ itself as a control time-series (as a covariate within $x$), will I recover the local linear trend model?



Your Answer

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

Browse other questions tagged or ask your own question.