hi @fabsig

From the readme,

https://github.com/fabsig/GPBoost#gpboost-and-lagaboost-algorithms

Focusing on the specific case of gaussian outcomes, Using the convention

y = F(X) + Zb + xi

where xi is an independent error term and X are predictor variables (aka covariates or features).

Is it possible to enable a feature where the user could supply in the values of xi. Therefore xi is not estimated, the user passes in a vector of values 'xi = c(1, ...., 10)' where the length of xi is equal to the length of the outcome

as,

# in R

gp_model <- fitGPModel(group_data=group_data, 
               xi=c(1,3,4, ... , 10) ,
               likelihood="gaussian",
               y=outcome_xgb, X=X)

If you think this is do-able, I could elaborate further on the motivation

Thank you.