Angie.H Moon

Conditional not onditional in the following https://github.com/betanalpha/knitr_case_studies/blob/25275128b6436b21662fc1b204c8733f46c357d5/stochastic_differential_equations/stochastic_differential_equations.Rmd#L344

Support InferenceData and Stanify

7

@jandraor's paper on applying Bayesian workflow could be extended with SBC. Currently system dynamics model translators are progressed in parallel ([python](https://github.com/hyunjimoon/pysd/tree/stan-backend) and [R](https://github.com/jandraor/readsdr/tree/sd_bayes) in the linked branch of the library)....

apply self-calib to precondition prior

34

We need prior and posterior to be similar ranges for valid testing (e.g. power). For two parameter normal models, tail area of scale papermeter is honed after over twenty iteration....

Inspection priority for High dimension

3

Preprocess tools that sort which parameter's rank plot to inspect first "high-dimensional models there are often only a few parameters/summaries of interest to the final application, and SBC is much...

Hierarchical model in complex models (nonlinear or nonparameteric or implicit)

1

1. With the help of this [code](https://github.com/hyunjimoon/defense-reliability/tree/master/R/2_Theta_bar_Y/spline/models), SBC-test spline model 2. With the established process, compare two hierarchical models - [gaussian process](https://github.com/hyunjimoon/defense-reliability/blob/master/vignettes/stancon21/sc21_notebook.Rmd) and spline model 3. Give example recommendations on...

Due to ADVI returning NULL, some results are NULL. NULL returns should be dropped and instead filled with the new attempt's nonNULL result. ``` > result$fits [[1]] NULL [[2]] variable...

focusSBC via information supplied from posteriordb or brms

1

Do SBC on partial joint probability space with some parameter values fixed. This would weigh down the needed computation greatly. From many SBC literature, researchers are interested in only certain...

1. Increase the number of num_warmup from M to at least 500 2. Only warm-up for a large number of samples at first. For the next step (N=2..), we could...

Nonlinearly increasing computation time

6

S = 10 ends within 1 minute, but with 100 it doesn't seem to finish (for over an hour) in `result_tn_100

chisq test pvalue as a measure

1

using pvalue as a global summary could be misleading as the following example suggests: ``` a