# \#loo

**URL:** https://discourse.mc-stan.org/tag/loo/33.md

[Latest](https://discourse.mc-stan.org/latest.md) · [Categories](https://discourse.mc-stan.org/categories.md) · [Tags](https://discourse.mc-stan.org/tags.md)

---

## [On LOO CV for non-factorizable normal models](https://discourse.mc-stan.org/t/on-loo-cv-for-non-factorizable-normal-models/7336)

<div class="topic-metadata">

**Author:** [@Elizaveta\_Semenova](https://discourse.mc-stan.org/u/Elizaveta_Semenova)\
**Replies:** 2\
**Last updated:** [September 17, 2026, 8:39pm UTC](https://discourse.mc-stan.org/t/on-loo-cv-for-non-factorizable-normal-models/7336 "2026-09-17T20:39:45Z")

</div>

Dear all, I have a question concerning LOO-CV for non-factorisable normal models, developed in the paper “Leave-one-out cross-validation for non-factorizable normal models” by Bürkner, Gabry, Vehtari. The idea of the pa…

---

## [Mixis and E\_loo](https://discourse.mc-stan.org/t/mixis-and-e-loo/40602)

<div class="topic-metadata">

**Author:** [@gregd](https://discourse.mc-stan.org/u/gregd)\
**Replies:** 25\
**Last updated:** [September 9, 2026, 11:41am UTC](https://discourse.mc-stan.org/t/mixis-and-e-loo/40602 "2026-09-09T11:41:58Z")

</div>

hello E\_loo uses the weights generated by loo to estimate the leave-one-out mean (and quantiles). However, “A small Pareto k estimate is necessary, but not sufficient, for E\_loo() to give reliable estimates.” Mixis is m…

---

## [Discretizing a hurdle model in order to compare it to a gaussian model: Part 2](https://discourse.mc-stan.org/t/discretizing-a-hurdle-model-in-order-to-compare-it-to-a-gaussian-model-part-2/41562)

<div class="topic-metadata">

**Author:** [@llewmills](https://discourse.mc-stan.org/u/llewmills)\
**Replies:** 4\
**Last updated:** [September 7, 2026, 9:31pm UTC](https://discourse.mc-stan.org/t/discretizing-a-hurdle-model-in-order-to-compare-it-to-a-gaussian-model-part-2/41562 "2026-09-07T21:31:40Z")

</div>

In this post a few days ago I attempted to get some advice on the procedure for discretizing a hurdle model and a gaussian model of the same outcome data so as to compare the two (recommended by @avehtari in his reply he…

---

## [LOO with brms::add\_criterion crashes the R session](https://discourse.mc-stan.org/t/loo-with-brms-add-criterion-crashes-the-r-session/41554)

<div class="topic-metadata">

**Author:** [@Anoban](https://discourse.mc-stan.org/u/Anoban)\
**Replies:** 5\
**Last updated:** [September 2, 2026, 1:22pm UTC](https://discourse.mc-stan.org/t/loo-with-brms-add-criterion-crashes-the-r-session/41554 "2026-09-02T13:22:47Z")

</div>

Hi folks, I’m trying to compare 8 brms models (6 as of now, still waiting for two to finish) using loo with moment\_match = TRUE but keep constantly running into crashes where the OS kills the R process. These models are…

---

## [LOO Continuously Ranked Probability Score](https://discourse.mc-stan.org/t/loo-continuously-ranked-probability-score/35239)

<div class="topic-metadata">

**Author:** [@cwolfe](https://discourse.mc-stan.org/u/cwolfe)\
**Replies:** 4\
**Last updated:** [August 11, 2026, 2:33pm UTC](https://discourse.mc-stan.org/t/loo-continuously-ranked-probability-score/35239 "2026-08-11T14:33:55Z")

</div>

I am comparing 2 models using the LOO implementation of continuously ranked probability scores. I just want to ensure I am interpreting the results correctly. It seems in most circumstances in applied research the CRPS i…

---

## [Difficulty adding loo criterion to longitudinal truncated normal model](https://discourse.mc-stan.org/t/difficulty-adding-loo-criterion-to-longitudinal-truncated-normal-model/41447)

<div class="topic-metadata">

**Author:** [@llewmills](https://discourse.mc-stan.org/u/llewmills)\
**Replies:** 14\
**Last updated:** [July 24, 2026, 8:44am UTC](https://discourse.mc-stan.org/t/difficulty-adding-loo-criterion-to-longitudinal-truncated-normal-model/41447 "2026-07-24T08:44:14Z")

</div>

I am analysing data from a clinical trial comparing placebo to active medication at multiple timepoint. One of my outcomes has a pronounced right-skew. The overwhelming majority of observations are between 0 and 0.03 but…

---

## [Model selection using ELPD LOOIC](https://discourse.mc-stan.org/t/model-selection-using-elpd-looic/41434)

<div class="topic-metadata">

**Author:** [@katiec](https://discourse.mc-stan.org/u/katiec)\
**Replies:** 4\
**Last updated:** [July 18, 2026, 8:14pm UTC](https://discourse.mc-stan.org/t/model-selection-using-elpd-looic/41434 "2026-07-18T20:14:55Z")

</div>

Hi! I want to assess the effects of first- and second-order temporal trends (t) and survey timing (doy) on my binomial response variable (occ = territory occupancy). The goal is to select the “best” baseline model to us…

---

## [Multilevel multiresponse phylogenetic mixed models with brms](https://discourse.mc-stan.org/t/multilevel-multiresponse-phylogenetic-mixed-models-with-brms/41411)

<div class="topic-metadata">

**Author:** [@Anoban](https://discourse.mc-stan.org/u/Anoban)\
**Replies:** 4\
**Last updated:** [July 10, 2026, 2:48am UTC](https://discourse.mc-stan.org/t/multilevel-multiresponse-phylogenetic-mixed-models-with-brms/41411 "2026-07-10T02:48:26Z")

</div>

Hi folks! I’m a 2nd year PhD student studying root ecological strategies and mycorrhizal outsourcing. I’m interested in understanding how the evolution of three response variables (three root traits) is influenced by a …

---

## [New loo package CRAN release v.2.10.0](https://discourse.mc-stan.org/t/new-loo-package-cran-release-v-2-10-0/41400)

<div class="topic-metadata">

**Author:** [@avehtari](https://discourse.mc-stan.org/u/avehtari)\
**Replies:** 3\
**Last updated:** [June 30, 2026, 8:13pm UTC](https://discourse.mc-stan.org/t/new-loo-package-cran-release-v-2-10-0/41400 "2026-06-30T20:13:35Z")

</div>

There is a new loo package CRAN release v2.10.0 with the most important new feature being new loo\_compare() output with columns p\_worse, diag\_diff and diag\_elpd, based on paper “Uncertainty in Bayesian leave-one-out cro…

---

## [How to handle high Pareto k when observations are meaningful?](https://discourse.mc-stan.org/t/how-to-handle-high-pareto-k-when-observations-are-meaningful/41368)

<div class="topic-metadata">

**Author:** [@cMiIIer](https://discourse.mc-stan.org/u/cMiIIer)\
**Replies:** 8\
**Last updated:** [June 18, 2026, 8:49pm UTC](https://discourse.mc-stan.org/t/how-to-handle-high-pareto-k-when-observations-are-meaningful/41368 "2026-06-18T20:49:34Z")

</div>

Hello, I am attempting to fit discrete time event models to some frog tadpole metamorphosis data. I have been able to successfully fit the model, but LOO for model comparisons demonstrate high pareto K values for rough…

---

## [Loo package support for Uncertainty in Bayesian leave-one-out cross-validation based model comparison](https://discourse.mc-stan.org/t/loo-package-support-for-uncertainty-in-bayesian-leave-one-out-cross-validation-based-model-comparison/41218)

<div class="topic-metadata">

**Author:** [@avehtari](https://discourse.mc-stan.org/u/avehtari)\
**Replies:** 0\
**Last updated:** [May 13, 2026, 5:59pm UTC](https://discourse.mc-stan.org/t/loo-package-support-for-uncertainty-in-bayesian-leave-one-out-cross-validation-based-model-comparison/41218 "2026-05-13T17:59:49Z")

</div>

In “Uncertainty in Bayesian leave-one-out cross-validation based model comparison” https://doi.org/10.1214/25-BA1569 we showed when LOO-CV elpd\_diff and se\_diff normal approximation uncertainty quantification in model co…

---

## [P\_loo \>\> p but no pareto-k warnings](https://discourse.mc-stan.org/t/p-loo-p-but-no-pareto-k-warnings/40837)

<div class="topic-metadata">

**Author:** [@ar2d2](https://discourse.mc-stan.org/u/ar2d2)\
**Replies:** 9\
**Last updated:** [April 27, 2026, 4:50pm UTC](https://discourse.mc-stan.org/t/p-loo-p-but-no-pareto-k-warnings/40837 "2026-04-27T16:50:33Z")

</div>

Hi, I’ve got a decently large RI-CLPM (N = 7796) that has Rhats \< 1.01, does not have pareto k warnings, but has a p\_loo \>\> p (see below). This concerns me since the loo glossary indicates that this can be a sign of sev…

---

## [Log Likelihood for Structural Equation Modeles](https://discourse.mc-stan.org/t/log-likelihood-for-structural-equation-modeles/41050)

<div class="topic-metadata">

**Author:** [@cMiIIer](https://discourse.mc-stan.org/u/cMiIIer)\
**Replies:** 2\
**Last updated:** [April 21, 2026, 8:24pm UTC](https://discourse.mc-stan.org/t/log-likelihood-for-structural-equation-modeles/41050 "2026-04-21T20:24:26Z")

</div>

Hello! I am working on a Dynamic Structural Equation model and I am having trouble generating log likelihood values for model comparisons. I am new to SEM in Stan and I am lost about how to proceed, I tried doing multid…

---

## [Add\_criterion applied to a model produced by brm()](https://discourse.mc-stan.org/t/add-criterion-applied-to-a-model-produced-by-brm/41037)

<div class="topic-metadata">

**Author:** [@JKingston](https://discourse.mc-stan.org/u/JKingston)\
**Replies:** 3\
**Last updated:** [April 9, 2026, 3:26pm UTC](https://discourse.mc-stan.org/t/add-criterion-applied-to-a-model-produced-by-brm/41037 "2026-04-09T15:26:43Z")

</div>

Please share your Stan program and accompanying data if possible. I am trying to use add\_criterion() to add “loo” to a model created with brm(), so that I can use loo\_compare() to compared models. add\_criterion() appea…

---

## [LOO and reinforcement learning models](https://discourse.mc-stan.org/t/loo-and-reinforcement-learning-models/41035)

<div class="topic-metadata">

**Author:** [@CamilleGrande](https://discourse.mc-stan.org/u/CamilleGrande)\
**Replies:** 4\
**Last updated:** [April 9, 2026, 11:24am UTC](https://discourse.mc-stan.org/t/loo-and-reinforcement-learning-models/41035 "2026-04-09T11:24:36Z")

</div>

Hi! Sorry if this is a very basic question, this is my first modelling project :) I am fitting reinforcement learning models for the Iowa Gambling Task using hBayesDM to real-world data. I followed the steps outlined in…

---

## [Kfolds with upper-level unit with brms model](https://discourse.mc-stan.org/t/kfolds-with-upper-level-unit-with-brms-model/41009)

<div class="topic-metadata">

**Author:** [@johezz](https://discourse.mc-stan.org/u/johezz)\
**Replies:** 1\
**Last updated:** [March 27, 2026, 7:57am UTC](https://discourse.mc-stan.org/t/kfolds-with-upper-level-unit-with-brms-model/41009 "2026-03-27T07:57:20Z")

</div>

Hi all, Does anyone know how I can do kfolds cross-validation by leaving one upper-level unit out instead of each observation in a brms model fitted on a hierarchical dataset? loo currently does the latter but not the f…

---

## [Memory issues when computing LOO/WAIC](https://discourse.mc-stan.org/t/memory-issues-when-computing-loo-waic/32576)

<div class="topic-metadata">

**Author:** [@jonas\_m](https://discourse.mc-stan.org/u/jonas_m)\
**Replies:** 6\
**Last updated:** [March 1, 2026, 3:52pm UTC](https://discourse.mc-stan.org/t/memory-issues-when-computing-loo-waic/32576 "2026-03-01T15:52:26Z")

</div>

Hi! Like several other people on this forum I am running into memory issues when trying to compute LOO/WAIC for large datasets. My data has roughly 160,000 observations. The model has three grouping levels, one with 15…

---

## [Are stacking weights “leaking” information across folds?](https://discourse.mc-stan.org/t/are-stacking-weights-leaking-information-across-folds/40892)

<div class="topic-metadata">

**Author:** [@Federico\_Billeci](https://discourse.mc-stan.org/u/Federico_Billeci)\
**Replies:** 7\
**Last updated:** [February 10, 2026, 2:28am UTC](https://discourse.mc-stan.org/t/are-stacking-weights-leaking-information-across-folds/40892 "2026-02-10T02:28:59Z")

</div>

Hey everyone, I have a bayesian model that can be written as p(y|x, \\theta ; t ) and it is basically a logistic regression, \\theta are the bayesian parameters and t is a hyperparameter. Instead of fixing t I wanted to f…

---

## [Model comparison for choice in per-subject latent-state models with additional measurement channels (LOO vs two-stage approach)](https://discourse.mc-stan.org/t/model-comparison-for-choice-in-per-subject-latent-state-models-with-additional-measurement-channels-loo-vs-two-stage-approach/40839)

<div class="topic-metadata">

**Author:** [@Anastasios\_D](https://discourse.mc-stan.org/u/Anastasios_D)\
**Replies:** 3\
**Last updated:** [January 22, 2026, 8:07pm UTC](https://discourse.mc-stan.org/t/model-comparison-for-choice-in-per-subject-latent-state-models-with-additional-measurement-channels-loo-vs-two-stage-approach/40839 "2026-01-22T20:07:21Z")

</div>

Hi everyone, I’m fitting per-subject Bayesian latent-state models in Stan for a gambling task (153 trials per participant). Each participant is fit independently. At each trial, a latent emotional state Et evolves acco…

---

## [Recent paper: Distributional bias compromises leave-one-out cross-validation](https://discourse.mc-stan.org/t/recent-paper-distributional-bias-compromises-leave-one-out-cross-validation/40743)

<div class="topic-metadata">

**Author:** [@rtnliqry](https://discourse.mc-stan.org/u/rtnliqry)\
**Replies:** 1\
**Last updated:** [December 15, 2025, 11:57am UTC](https://discourse.mc-stan.org/t/recent-paper-distributional-bias-compromises-leave-one-out-cross-validation/40743 "2025-12-15T11:57:22Z")

</div>

Hi Stan forum, An interesting paper was published recently in Science Advances that details how LOO-CV can introduce a (typically small) negative bias to model performance estimates. They propose a “rebalanced LOO-CV” t…

---

## [LOO in Meta-analysis with mixed aggregate and individual data](https://discourse.mc-stan.org/t/loo-in-meta-analysis-with-mixed-aggregate-and-individual-data/40363)

<div class="topic-metadata">

**Author:** [@timdisher](https://discourse.mc-stan.org/u/timdisher)\
**Replies:** 6\
**Last updated:** [September 18, 2025, 5:13pm UTC](https://discourse.mc-stan.org/t/loo-in-meta-analysis-with-mixed-aggregate-and-individual-data/40363 "2025-09-18T17:13:09Z")

</div>

I’ve been experimenting with the multinma package for multi-level network meta-regression (ML-NMR), and I’ve run into something that’s left me unsure how to interpret LOO-CV results in this setting. In the multinma pack…

---

## [Getting inscrutable error when running add\_criterion() on brms ordinal regression model](https://discourse.mc-stan.org/t/getting-inscrutable-error-when-running-add-criterion-on-brms-ordinal-regression-model/40326)

<div class="topic-metadata">

**Author:** [@llewmills](https://discourse.mc-stan.org/u/llewmills)\
**Replies:** 10\
**Last updated:** [September 18, 2025, 8:04am UTC](https://discourse.mc-stan.org/t/getting-inscrutable-error-when-running-add-criterion-on-brms-ordinal-regression-model/40326 "2025-09-18T08:04:26Z")

</div>

I am conducting a multivariate ordinal regression examining the effect of number of days of amphetamine-type-stimulant use in the 28 days prior to starting treatment (variable ats\_baseline) on three outcomes: psychologic…

---

## [High Pareto k̂ values only for some datapoints for some subjects](https://discourse.mc-stan.org/t/high-pareto-k-values-only-for-some-datapoints-for-some-subjects/40234)

<div class="topic-metadata">

**Author:** [@Hadeel\_Haj\_Ali](https://discourse.mc-stan.org/u/Hadeel_Haj_Ali)\
**Replies:** 6\
**Last updated:** [September 2, 2025, 12:58pm UTC](https://discourse.mc-stan.org/t/high-pareto-k-values-only-for-some-datapoints-for-some-subjects/40234 "2025-09-02T12:58:17Z")

</div>

I’m working on a task where participants make risky choices on a continuous scale (0 to 50) across 20 trials. My model includes standard risk-related parameters and allows them to drift over trials so each parameter has …

---

## [Saving log-likelihood with parallelized implementation](https://discourse.mc-stan.org/t/saving-log-likelihood-with-parallelized-implementation/40263)

<div class="topic-metadata">

**Author:** [@Loni92](https://discourse.mc-stan.org/u/Loni92)\
**Replies:** 1\
**Last updated:** [August 22, 2025, 6:47pm UTC](https://discourse.mc-stan.org/t/saving-log-likelihood-with-parallelized-implementation/40263 "2025-08-22T18:47:44Z")

</div>

Hi all, I am trying to implement a parallelized version of my model using the reduce\_sum function. I am wondering if it is somehow possible to still save the log-likelihood of each data point for model comparison later …

---

## [Comparing brms Univariate vs. Joint Models using PSIS-LOO. Is ELPD additive?](https://discourse.mc-stan.org/t/comparing-brms-univariate-vs-joint-models-using-psis-loo-is-elpd-additive/40112)

<div class="topic-metadata">

**Author:** [@willashton](https://discourse.mc-stan.org/u/willashton)\
**Replies:** 4\
**Last updated:** [August 17, 2025, 2:52pm UTC](https://discourse.mc-stan.org/t/comparing-brms-univariate-vs-joint-models-using-psis-loo-is-elpd-additive/40112 "2025-08-17T14:52:02Z")

</div>

Hi all! First time poster, long time struggler :) I am fairly new to Bayesian modelling (though I’ll be starting a PhD in it come October). I’m currently working on a modelling project using brms for my MSc dissertatio…

---

## ["Hypothesis testing" with multi-level models (for a Registered Report)](https://discourse.mc-stan.org/t/hypothesis-testing-with-multi-level-models-for-a-registered-report/40037)

<div class="topic-metadata">

**Author:** [@Alasdair\_Clarke](https://discourse.mc-stan.org/u/Alasdair_Clarke)\
**Replies:** 8\
**Last updated:** [August 12, 2025, 12:47pm UTC](https://discourse.mc-stan.org/t/hypothesis-testing-with-multi-level-models-for-a-registered-report/40037 "2025-08-12T12:47:07Z")

</div>

Hello, I am helping a friend with a Registered Report (a paper format inwhich your introduction + methods are peer reviewed before you conduct the experiment). As part of this, we’re working out our decision criteria an…

---

## [MCSE of elpd\_loo is NA! How to proceed?](https://discourse.mc-stan.org/t/mcse-of-elpd-loo-is-na-how-to-proceed/40097)

<div class="topic-metadata">

**Author:** [@arumadri](https://discourse.mc-stan.org/u/arumadri)\
**Replies:** 2\
**Last updated:** [August 11, 2025, 2:53pm UTC](https://discourse.mc-stan.org/t/mcse-of-elpd-loo-is-na-how-to-proceed/40097 "2025-08-11T14:53:22Z")

</div>

Hi there, I’m trying to fit a model including crossbasis matrices generated using Distributed Lag Non-linear Models (dlnm) as predictors using brms (this has been solved/fitted using INLA elsewhere here). My goal is to …

---

## [Model selection with multiple observation likelihoods](https://discourse.mc-stan.org/t/model-selection-with-multiple-observation-likelihoods/39945)

<div class="topic-metadata">

**Author:** [@jkbest2](https://discourse.mc-stan.org/u/jkbest2)\
**Replies:** 3\
**Last updated:** [July 29, 2025, 11:59am UTC](https://discourse.mc-stan.org/t/model-selection-with-multiple-observation-likelihoods/39945 "2025-07-29T11:59:05Z")

</div>

I’m looking to do model selection with the loo package on a mark-recapture model that I’ve been working on. The experimental setup is fairly simple. Fish are captured in a trap, marked, and released upstream of the trap.…

---

## [Psis-loo with censored observations](https://discourse.mc-stan.org/t/psis-loo-with-censored-observations/27252)

<div class="topic-metadata">

**Author:** [@chvandorp](https://discourse.mc-stan.org/u/chvandorp)\
**Replies:** 12\
**Last updated:** [July 25, 2025, 2:32pm UTC](https://discourse.mc-stan.org/t/psis-loo-with-censored-observations/27252 "2025-07-25T14:32:40Z")

</div>

Hi all, although I found a very similar thread on the forum (Loo: calculating point-wise log-lik when using data augmentation for censored observations), I don’t think my question has been answered. Consider the followi…

---

## [Question on model stacking -- averaging posteriors of parameters, not predictions](https://discourse.mc-stan.org/t/question-on-model-stacking-averaging-posteriors-of-parameters-not-predictions/21313)

<div class="topic-metadata">

**Author:** [@tvladeck](https://discourse.mc-stan.org/u/tvladeck)\
**Replies:** 19\
**Last updated:** [July 24, 2025, 6:34am UTC](https://discourse.mc-stan.org/t/question-on-model-stacking-averaging-posteriors-of-parameters-not-predictions/21313 "2025-07-24T06:34:15Z")

</div>

The situation is as follows: We are running a large model that takes quite a long time to run. It takes too long to realistically use the the approach in the “leave-future-out” vignette It includes a GP whose kernel h…

[Next page](https://discourse.mc-stan.org/tag/loo/33.md?match_all_tags=true&page=1&tags%5B%5D=loo)