Measurement Uncertainties in Multiple Regression

Thanks!

I’d parsed out the variables because in my real case some of the uncertainties have different distributions, but your formulation is definitely more elegant!

Out of interest, is there a reason you calculate the likelihood using mu - y_true rather than using mu and setting observed=y_true? I’d always assumed that setting observed was necessary to tell the model what the target was… how would this model know what data it’s fitting?