Intro
Contents
Intro¶
Regression examples illustrating Bayes window, a suite of classes for onle-liner estimation of posteriors and linear mixed models.
Why estimation statistics¶
Visual representation of confidence interval
Allows to eyeball effect size
Easy to explain to non-scientists
Hypothesis testing-free (Except LME)
Non-standard distributions
Count data (eg action potentials of neurons)
Lognormal effects (common in many fields)
Why bayes-window¶
Compact envocation
Bayesian models up to hiearchical regression with just a single call
Faceted visualization included
API free of Bayesian jargon
Robust visualization with or without fitting a model
Baked-in overlay of model output onto exploratory plots (eg boxplot)
Easy to see when model fails to capture data
Composable graphs (eg lme_ci_graph | bayes_hdi_with_boxplot)
Why not arviz?¶
No faceting
No integration with model’s intention
For a meaningful pub-ready presentation of posteriors over even one type of condition, arviz output is only usable after at least a page of code