Closed stevencarlislewalker closed 2 years ago
If the user wants to express time variation as a dependence on past values (rel_prev or rel_orig), then normal multiplication does not work well for parameters that must lie between zero and one.
What we need is a logit-product:
x * y = plogis(qlogis(x) + qlogis(y))
If the user wants to express time variation as a dependence on past values (rel_prev or rel_orig), then normal multiplication does not work well for parameters that must lie between zero and one.
What we need is a logit-product:
x * y = plogis(qlogis(x) + qlogis(y))