search

fabsig / GPBoost

Combining tree-boosting with Gaussian process and mixed effects models
Other
571 stars 46 forks source link

dependent random effects #144

Closed avivihadar closed 4 months ago

avivihadar commented 4 months ago

Hi,

Thank you so much for this package.

Could you allow the random effects to depend on each other (i.e., don't enforce the covariances in the covariance matrices to be zero)?

Best, Hadar

fabsig commented 4 months ago

Thank you for your interest in GPBoost!

When you use Gaussian process-based random effects, the random effects are correlated (e.g. over space and/or time). For grouped random effects, dependence is currently not supported in GPBoost, i.e., it is assumed that they are independent a priori (given the data, they can be dependent). Except for random slopes, dependence does not make a lot of sense anyway for, e.g., crossed or nested random effects.