OK, I’ve started the original on Colab:
Things aren’t that great here either apparently:
The rhat statistic is larger than 1.4 for some parameters. The sampler did not converge.
The estimated number of effective samples is smaller than 200 for some parameters.
(would be great if Turing gave similar commentary…)
Latest traces are:
{'p0_logodds__': -1.1201974288020848, 'p1_logodds__': -1.8108031775461457, 'N': 94, 'p0': 0.24597466437185683, 'p1': 0.1405410824792338}
{'p0_logodds__': -1.0759843501248898, 'p1_logodds__': -1.3370046772297803, 'N': 95, 'p0': 0.25426669304142646, 'p1': 0.208003069852844}
{'p0_logodds__': -1.309490295895163, 'p1_logodds__': -2.229973878615384, 'N': 97, 'p0': 0.21257214856225384, 'p1': 0.09709093087482638}
{'p0_logodds__': -1.2321949553810507, 'p1_logodds__': -2.4305296251833313, 'N': 96, 'p0': 0.22579748811056297, 'p1': 0.0808740895970291}
which aren’t far from the initial parameters I’ve tried (p0=0.2, p1=0.2, N=100).
I’ve tried:
trace9 = sample(model9(k01, k10, k11), Gibbs(NUTS(1000,.8,:p0,:p1),PG(20,:N)), 1000, init_params=[0.22, 0.08, 95])
But no joy:
