replace "sampling statement" with "target augmentation statement"

[bob-carpenter]

Summary:

The ~ notation continues to trip up users.

Description:

See summary.

Current Version:

v2.20something

[bob-carpenter]

@avehtari commented via email: I think we should emphasize the common meaning of the tilde in statistics "has the distribution (of)", for example, y ~ normal(mu, sigma) should be read as "y is distributed as normal(mu, sigma)".

I like this better, but it's still potentially confusing in that if we write theta ~ uniform(0, 1) for a parameter theta, it doesn't have a uniform(0, 1) distribution in the posterior.

[bob-carpenter]

@andrewgelman responded via email: I would not want to write it as "y is distributed as normal(mu, sigma)" because consider this example:

theta ~ normal(0, 1);
theta ~ normal(0, 1);

for which it is not true that theta has a normal(0, 1) distribution!