Hi, While building models and trying out different priors, I noticed that using a prior that seemed more sensible based on my sensitivity analysis decreased my loo performance. The old prior, the one before the sensitivity analysis, also seems to sample slightly better. Would it be correct to go w…

[image] avehtari: Can you post some figures with explanations how to interpret sensitivity from those? I started with this prior as a starting point: [psa] My interpretation of this was, that most of the curves were too steep. For most of the parameter space a unit chance in LOO would have n…

Balancing loo performance and Prior sensitivity analysis

Modeling

avehtari August 3, 2019, 6:29pm 10

Excellent figures. It seems the mystery is now solved?

1 Like

Brms prior more influential than data itself

Topic		Replies	Views
Loo, model_weights and brms models with sample_prior = "only" Modeling loo , brms	3	435	December 11, 2020
LOO error with a brmsfit model brms loo	3	482	October 2, 2018
Loo_compare() to compare models with different priors General	2	215	December 1, 2023
Variational approximation for LKJ from Giordano et al Developers	2	682	January 12, 2017
Checking model and LOO with models of different observations Modeling fitting-issues , performance , loo , brms	4	425	May 7, 2021

Balancing loo performance and Prior sensitivity analysis

Related Topics