/* Function to call genotypes from post probs
NOTE: probs must be normalized, that is, sum (or logsum) to 1!
geno : array with probs
n_geno : number of genotypes in array
log_scale : are the probs in log scale?
N_prob_thresh : minimum prob to use data.
If highest prob is lower, data set as missing
call_prob_thresh : minimum prob to call a genotype.
If highest prob is lower, leave geno as probs
miss_data : how the missing data is handled
0 = missing data (all genot with equal prob)
1 = sample random genotype
2 = call the highest prob geno (since missing data, probably major/major)
*/
void call_geno(double *geno, int n_geno, bool log_scale, double N_prob_thresh, double call_prob_thresh, int miss_data){
if(N_prob_thresh > call_prob_thresh)
error(__FUNCTION__, "missing data threshold must be smaller than calling genotype threshold!");
int max_pos = array_max_pos(geno, n_geno);
int min_pos = array_min_pos(geno, n_geno);
double max_pp = (log_scale ? exp(geno[max_pos]) : geno[max_pos]);
// If missing data
if(geno[min_pos] == geno[max_pos]){
if(miss_data == 0)
max_pp = -1;
else if(miss_data == 1)
max_pos = rand() % 3;
}
if(max_pp < N_prob_thresh)
for (int g = 0; g < n_geno; g++)
geno[g] = (log_scale ? log((double) 1/n_geno) : (double) 1/n_geno);
if(max_pp >= call_prob_thresh){
for (int g = 0; g < n_geno; g++)
geno[g] = (log_scale ? -INF : 0);
geno[max_pos] = (log_scale ? log(1) : 1);
}
}
void call_geno(double *site_gl, int n_ind, int n_geno) {
for (int i = 0; i < n_ind; i++) {
int max_pos = array_max_pos(&site_gl[i*n_geno], n_geno);
for (int j=0; j < n_geno; j++) site_gl[i*n_geno+j] = -1e4;
site_gl[i*n_geno+max_pos] = 0;
}
}
