Интересно, что в отличие от авторов-человечков, гуглевский ИИ сразу смекнул, что мне надо и зачем:
Цитата
Zero-intercept Poisson regression is a specialized form of Poisson regression where the intercept term is constrained to be zero. This means that when all predictor variables are zero, the predicted count is also zero. It's used when there's a theoretical or empirical reason to believe that the relationship between the predictors and the outcome should pass through the origin.
When is it appropriate?
Prior Knowledge:
If you have strong prior knowledge or a theoretical model that dictates the outcome should be zero when all predictors are zero, a zero-intercept model is justified.
When is it appropriate?
Prior Knowledge:
If you have strong prior knowledge or a theoretical model that dictates the outcome should be zero when all predictors are zero, a zero-intercept model is justified.
Но, увы, никаких конкретных реализаций это-самой Zero-intercept Poisson regression, найти мне не удалось, везде zero-inflated да zero-truncated толкают, что совсем не о том.
