search

X-DataInitiative / tick

Module for statistical learning, with a particular emphasis on time-dependent modelling
https://x-datainitiative.github.io/tick/
BSD 3-Clause "New" or "Revised" License
480 stars 105 forks source link

Plot convention question #473

Closed Jbollenbacher closed 3 years ago

Jbollenbacher commented 3 years ago

The plots produced by plot_hawkes_kernels are ambiguous. Does a peak in kernel(0,1) indicate that events in node 0 precede events in node 1, or that events in node 1 precede events in node 0? There is no documentation of this convention, as far as I can tell.

For instance, in the below plot of results from a HawkesSumGaussians fit, I cant tell if 1 causes 0 or 0 causes 1. Which is it?

Thanks

Screen Shot 2021-04-27 at 5 50 00 PM
Jbollenbacher commented 3 years ago

I'm assuming it follows the same convention as Phi_(i,j) here, but the documentation doesn't say so explicitly.

Bacry commented 3 years ago

Yes, Phi_(i,j) is the kernel that codes the influence of Past events on the jth component to present intensity of the ith component