the Stan language allows vector and matrix declarations:
matrix[0,0] m0;
vector[0] v0;
row_vector[0] rv0;
would objects like the above ever be used?
the Stan language allows vector and matrix declarations:
matrix[0,0] m0;
vector[0] v0;
row_vector[0] rv0;
would objects like the above ever be used?
Yes. Typically not declared with the literal 0, but usually with variables.
It’s a quick way to sample from the prior. Or have 0-length observations
for a group.
Yeah we do what Daniel mentioned a lot in the Stan programs for rstanarm. For example, it allows us to do something like
data {
...
int<lower=0,upper=1> has_intercept;
}
parameters {
...
real alpha[has_intercept];
}
so alpha
will only have positive length if the user has specified a model with an intercept.
I guess I used an array and not a vector in that example, but it’s the same idea.