Closed jingwen66 closed 4 years ago
No reason other than it hasn't been implemented yet due to lack of time
Is it possible to get it implemented in near future :)
Possibly but I'm not sure exactly when I'll next have time to work on it
Hi, I've implemented this p-value and I used the same technique described on the paper. `post_inf = impact.inferences[post_period[0]:post_period[1]] step_x = np.random.randint(2, size=len(post_inf.index) 1000) step_x[step_x == 0] = -1 x = post_inf['point_pred'].repeat(1000) + abs(post_inf['point_pred'].repeat( 1000) - post_inf['point_pred_upper'].repeat(1000)) np.random.random(size=len(post_inf.index) * 1000) aciertos = post_inf['response'].repeat(1000) - x > 0
goals_f = aciertos.sum() / (len(post_inf.index) * 1000)`
Hope this will be useful
@Worerlz You should submit a pull request so if it works it can be merged in to Master upon @jamalsenouci approval
Hi @jamalsenouci , just wondering, do you know how we can implement the computation of the p-value in this version of python? Maybe with some guidance we can contribute to the project.
Google's code uses an average over several samples, not sure how the same could be implemented here (or even if @Worerlz suggestion already does the trick)
I have attempted a p-value implementation in the latest version. This will need updating when the bayesian estimation is done but should do for now.
Summary of R package provides "Posterior tail-area probability p" and "Posterior prob. of a causal effect". I cannot find it in summary of python package. Is there any reason of not reporting them?