Hi all,

I am new to this field.

I try the tutorial Bayesian Imputation for Missing Values in Discrete Covariates (NumPyro documentation ).

I thought that the following code does not cancel the log_prob of the value that replaced the missing value.

Shouldn’t the missing value be cancelled or log_prob=0 as if it were a continuous value?

```
# cancel out enumerated values that are not equal to observed values
log_prob = jnp.where(A_isobs & (Aimp != A), -inf, log_prob)
```

I would be thankful if you could help me with this question.