Reconstructed demonstration · mock values
From uncertain factors to possible outcomes
Follow one synthetic batch through the model. Select a factor to trace its connections.
Possible outcomes
Dots: model means. Lines: central 90% ranges of modeled probabilities or expected efficiency, not intervals for observed results.
A distribution preserves the range of possible operating conditions.
Mock example only. No hospital data, fitted parameters, or original model inputs are shown. The network links are reconstructed from the project code; all numerical values are synthetic. Arrows represent model assumptions, not demonstrated intervention effects.
How this demonstration is calculated
Five independent beta priors generate 12,000 candidate factor sets. Fixed illustrative logistic coefficients connect them to three session-level binary outcomes and expected efficiency. A synthetic batch of 20 sessions supplies binomial likelihoods and a normal likelihood for mean efficiency. Importance weights approximate the joint posterior. The animation draws whole factor sets from that posterior, preserving the dependence introduced by conditioning on shared outcomes.
The original implementation used PyMC and MCMC to estimate uncertain factors, weights, and intercepts. This small demonstration fixes the coefficients and omits room, weekday, and staffing effects. The start-to-end link uses predicted start probability, as in the project code. It illustrates the mechanics, not the original model's fit or hospital performance.