This should work:
data{
int which_fixed ;
real fixed_value ;
… // other data
}
parameters{
vector[4] abcd ;
… // other parameters
}
model {
abcd[1] ~ … // prior for a
abcd[2] ~ … // prior for b
… // c & d priors too
vector[4] abcd_with_fixed = abcd ;
abcd_with_fixed[ which_fixed ] = fixed_value ;
… // other model structure
}
…
You’ll still get samples for the fixed parameter that represent samples from the prior, but everything else will reflect a posterior conditional on that parameter being fixed at the supplied value.