/*
* Calculate adjustment for missing data in pairwise comparisons. This will
* return a square matrix that is used to multiply the raw differences of a
* distance matrix in order to scale the differences by the number of observed
* loci.
*
* Parameters:
* nas a list where each element represents a sample containing an integer
* vector representing positions of missing data for that individual
* nloc an integer specifying the number of loci observed in the entire set
*
* Return:
* a square matrix
*/
SEXP adjust_missing(SEXP nas, SEXP nloc)
{
int i;
int j;
int NLOC = asInteger(nloc);
SEXP nai;
SEXP naj;
double u;
int n = length(nas);
SEXP out = PROTECT(allocMatrix(REALSXP, n, n));
for (i = 0; i < n - 1; i++)
{
// set diag to one
REAL(out)[i + i*n] = 1.0;
// GET NA list for i
nai = VECTOR_ELT(nas, i);
for (j = i + 1; j < n; j++)
{
// Get NA list for j
naj = VECTOR_ELT(nas, j);
// Scale by N/(N - M)
u = (double)NLOC/(double)(NLOC - count_unique(nai, naj));
REAL(out)[i + j*n] = u;
REAL(out)[i*n + j] = u;
}
}
// fencepost for identity
REAL(out)[(n*n) - 1] = 1.0;
UNPROTECT(1);
return(out);
}
void print_stat(int* data, int start, int end_exclusive) {//UNTESTED
int unique_values= count_unique(data, start, end_exclusive);
printf("[%d,%d,%d]\n",start,end_exclusive-1, unique_values);
}
inline std::size_t count_unique(Iter first, Iter last) {
using value_type = typename std::iterator_traits<Iter>::value_type;
return count_unique(first, last, std::equal_to<value_type>());
} // count_unique
