What is the difference between specifying X ~ N(mu, simga2) and target += normal_lpdf(mu, sigma2)

[jonah]

Turns out this is just due to the fact that rounding is impacted by whether the constants are included, which definitely didn’t occur to me but makes total sense:

Why does dropping constants give different parameter estimates if seed and inits are the same?

Look to see if the first couple iterations are the same. After that, whether there are or are not “constants” can cause it to round differently on the last decimal place and eventually cause the chains to differ.