Apologies if this is a silly question, but couldn’t find the answer.

I have a transactional dataset that I would like to model so that I can do an A/B test.

The model is:

`S(x) = T * V(x)`

where `T`

is the probability that a user buys something and `V(x)`

is the probability that a purchase will have log value `x`

`T`

can be modeled as `pm.Bernoulli`

and `V(x)`

can be modeled as a `pm.Normal`

Now I would have thought that one could do:

```
conv = pm.Bernoulli("conversion",p) #priors are defined somewhere else before the code
value = pm.Normal("value", mu, sd)
val = pm.Deterministic('name',conv*value, observed=data)
```

But that doesn’t work, as Deterministic doesn’t take `observed`

.

How do I specify a probability function that is the product of `conv`

and `value`

so that I can get samples of the posterior for `p`

and `mu`

?

Thanks.