NB support

[wangxinge]

Hi there, Can I use glm.nb for the regression? Thank you!

[lindeloev]

Negative Binomial regression is not natively supported yet, but I will make it a priority adding it. Specifically:

• [ ] Support for mcp(..., family = negbinomial()) with appropriate default priors.
• [ ] Allow changepoint regression on the shape, e.g., model = list(y ~ 1 + x + shape(1 + x), data = df)
• [ ] Add support for which_y = "shape" in plot(), fitted(), etc.

Until then, should be able to "hack" it by looking at the JAGS code in fit$jags_code for a regular binomial model and replace dbin(y_[i_], N[i_]) with dnegbin(y_[i_], shape) and add an appropriate prior, e.g., for shape ~ dunif(0, 50). Then feed that string into mcp again via mcp(model, data, jags_code = your_updated_code). This won't return posterior samples for shape, unfortunately.