Several of our algorithms seem to fail sporadically on this model and report an incorrect mean. The incorrect value is always roughly the same, 0.38 for the mean.
Running this locally, it seems that both smc and mala seem to often coverge to a number around 0.35, which makes me wonder two things:
First, is our analytic calculation of the mean correct? We currently have the value as 7/25 = 0.28
Second, is our testing tolerance too high? (almost certainly)
Several of our algorithms seem to fail sporadically on this model and report an incorrect mean. The incorrect value is always roughly the same,
0.38for the mean.Running this locally, it seems that both smc and mala seem to often coverge to a number around 0.35, which makes me wonder two things: First, is our analytic calculation of the mean correct? We currently have the value as 7/25 = 0.28
Second, is our testing tolerance too high? (almost certainly)
Failures
tempered_smc: https://github.com/flatironinstitute/bayes-kit/actions/runs/4008341002/jobs/6882368677 https://github.com/flatironinstitute/bayes-kit/actions/runs/4008269276/jobs/6882211668 https://github.com/flatironinstitute/bayes-kit/actions/runs/3971370594/jobs/6808183840 https://github.com/flatironinstitute/bayes-kit/actions/runs/3971574251/jobs/6808622237 https://github.com/flatironinstitute/bayes-kit/actions/runs/4000634814/jobs/6865992560
mala: https://github.com/flatironinstitute/bayes-kit/actions/runs/3971574251/jobs/6808622046