Error building wrapper around Rcpp function - r

Question

I have the following function declared in Rcpp:
#include <Rcpp.h>
// [[Rcpp::depends(RcppArmadillo)]]
#include <Rmath.h>
using namespace Rcpp;
// [[Rcpp::export]]
double loglikZeta(double zold, double zstar, NumericVector y, int K, double p){
NumericVector num = Rcpp::dbinom(y,K,p*zstar);
NumericVector den = Rcpp::dbinom(y,K,p*zold);
return (num[0]/den[0]);
}
// [[Rcpp::export]]
double singleZetaSampler(NumericVector z, NumericVector y,
double p, int K, int i, double zstar){
return loglikZeta(z[i-1],zstar,y[i-1],K,p);
}
Now declare (after having loaded package and file):
z <- y <- c(rep(1,20),rep(0,20))
n <- length(y)
K <- 3
p <- 0.5
i <- 30
zstar <- 1
The unexpected behaviour is that if I try to call I have everytime different results (there is nothing random in the function):
singleZetaSampler(z,y,p,K,i,zstar)
[1] 1.000051
singleZetaSampler(z,y,p,K,i,zstar)
[1] 0.1887447
singleZetaSampler(z,y,p,K,i,zstar)
[1] 0.9999998
Is there any big error am I doing here or these results are actually unexpected?
EDIT:
Sorry if the function doesn't make sense used as it is. This was the original function:
// [[Rcpp::export]]
NumericVector zetaSampler(int n, NumericVector z, NumericVector y,
double p, int K){
NumericVector xx(n);
for(int i = 0; i < n; i++){
xx(i) = loglikZeta(z[i],1,y[i],K,p);
}
return xx;
}
and calling:
zetaSampler(length(z),z,y,p,K)
as before gives different results every time.

Answer 1

Two things. One actual error, one sort-of-stylistic.
The stylistic issue is that you include Rmath.h and depend on RcppArmadillo when you should not. The real error is that you sample 20 times but then set i=30 and access the 30th element. So you get random inputs.
Here is what I just ran, and it gets three times the same result.
#include <Rcpp.h>
using namespace Rcpp;
// [[Rcpp::export]]
double loglikZeta(double zold, double zstar, NumericVector y, int K, double p){
NumericVector num = Rcpp::dbinom(y,K,p*zstar);
NumericVector den = Rcpp::dbinom(y,K,p*zold);
return (num[0]/den[0]);
}
// [[Rcpp::export]]
double singleZetaSampler(NumericVector z, NumericVector y,
double p, int K, int i, double zstar){
return loglikZeta(z[i-1],zstar,y[i-1],K,p);
}
/*** R
z <- y <- c(rep(1,20),rep(0,20))
n <- length(y)
K <- 3
p <- 0.5
i <- 20 # not 30
zstar <- 1
singleZetaSampler(z,y,p,K,i,zstar)
singleZetaSampler(z,y,p,K,i,zstar)
singleZetaSampler(z,y,p,K,i,zstar)
*/
Output:
R> sourceCpp("/tmp/foo.cpp")
R> z <- y <- c(rep(1,20),rep(0,20))
R> n <- length(y)
R> K <- 3
R> p <- 0.5
R> i <- 20 # not 30
R> zstar <- 1
R> singleZetaSampler(z,y,p,K,i,zstar)
[1] 1
R> singleZetaSampler(z,y,p,K,i,zstar)
[1] 1
R> singleZetaSampler(z,y,p,K,i,zstar)
[1] 1
R>
Edit: Appears to work better in a repaired version forcing scalar arguments to loglikZeta():
// [[Rcpp::export]]
double loglikZeta(double zold, double zstar, double y, int K, double p){
double num = R::dbinom(y, K, p*zstar, false);
double den = R::dbinom(y, K, p*zold, false);
return (num/den);
}
Note that Rcpp::dbinom() has a signature of Rcpp::dbinom(Rcpp::NumericVector, int, double, bool=false).