Accept or Reject
A hollow candidate opens along the parameter ladder; accepted proposals pull the filled state onto them and the rejected one recoils and fades while the state holds.
A hollow candidate opens along the parameter ladder; accepted proposals pull the filled state onto them and the rejected one recoils and fades while the state holds.
A pointer travels out from parity along a graduated evidence scale, pushed unevenly past the threshold marks by a growing evidence bar into decisive support for one model.
A cut rule walks out across a settling chain to the end of burn-in, the retained tail carrying the accent while the discarded head falls back to grey.
A sampling head sweeps the window and draws in three chains started from separated points, which lose their starting values and settle together into the shaded target band.
Successes and failures ratchet into two tally columns a cell at a time while the posterior mean bar steps up from the dashed prior mean to the ratio the counts imply.
Two caps open out across a discrete posterior and draw back in, the accented mass between them narrowing to the credible region while the tails stay grey.
Group estimates slide out to their unpooled tick marks and are drawn partway back toward the pooled mean rule, which never moves.
A scan line walks the parameter axis over a fixed strip of observations while a gauge reads the likelihood, overshooting either side before settling on the notched maximum.
A sampler trace scrolls steadily through a fixed window against the dashed target level, reading as a chain still running rather than a finished draw.
Three candidate models run their score bars out from a common origin a beat apart while the selection bracket closes from all three rows onto the winning one.
A broad off-centre prior density contracts onto the dashed data mean, growing taller and narrower for the same mass as the posterior resolves.
A bundle of replicated trajectories collapses and fans back out to its full predictive spread while the observed statistic holds still inside it.
A row of equal-height belief bars over hypothesis bins redistributes its mass toward the bins the evidence favours, leaving a concentrated posterior standing above the flat prior level.
A ladder of posterior intervals arrives batch by batch, each row narrower and closer to the fixed truth rule than the one above it.
Every third tick leaves the comb of draws and drops in sequence to the kept row below, while the discarded draws stay on the rail in grey.