NAN in log_marginal_likelihood

Questions v5
1

I constructed a model and sampled it through SMC to obtain the marginal distribution. But in the final sample_stats, the log_marginal_likelihood contains a lot of nan. In every chain, it only sampled 13 marginal likelihoods. I don’t know what caused this.

I am looking forward to your reply. Thanks very much!

log_marginal_likelihood
(chain, draw)
object
nan nan nan … nan nan -4.652e+04
array([[nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan,
-46518.09266605749],
[nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan,
-46518.07094694604],
[nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan,
-46518.05669340289],
[nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan,
-46518.077746311734]], dtype=object)

2

The nan are just padding if I recall correctly. Only the last values matter.