value gamma_CAML_rates( value a, value b, value c )
{
CAMLparam3(a,b,c);
CAMLlocal1( rates );
double alpha,beta,*rate_ray,*pcut_ray;
int cats,j;
long dims[1];
alpha = Double_val( a );
beta = Double_val( b );
cats = Int_val( c );
assert( cats > 0 );
rate_ray = (double*) malloc( sizeof(double)*cats );
CHECK_MEM(rate_ray);
if( 1 == cats ){
rate_ray[0] = 1.0;
} else {
pcut_ray = (double*) malloc( sizeof(double)*cats );
CHECK_MEM(pcut_ray);
for(j=1;j<cats;++j)
pcut_ray[j-1] = gamma_pp( (double)j/(double)cats, alpha, beta );
gamma_rates( rate_ray, alpha, beta, pcut_ray, cats );
free( pcut_ray );
}
dims[0] = cats;
rates = alloc_bigarray(BIGARRAY_FLOAT64 | BIGARRAY_C_LAYOUT,1,rate_ray,dims);
CAMLreturn( rates );
}
void
gamma_rate_mixture_summarize(
arb_ptr rate_mix_prior, arb_ptr rate_mix_rates, arb_ptr rate_mix_expect,
const gamma_rate_mixture_t x, enum rate_mixture_mode mode, slong prec)
{
slong i;
arb_ptr r;
arb_t p, q;
arb_init(p);
arb_init(q);
arb_set_d(p, x->invariable_prior);
arb_sub_si(q, p, 1, prec);
arb_neg(q, q);
/* the expectation is designed to be 1 */
arb_one(rate_mix_expect);
/* probabilities corresponding to gamma rate categories */
for (i = 0; i < x->gamma_categories; i++)
{
r = rate_mix_prior + i;
arb_div_si(r, q, x->gamma_categories, prec);
}
/* optional probability corresponding to the invariable category */
if (x->invariable_prior)
{
r = rate_mix_prior + x->gamma_categories;
arb_set(r, p);
}
/* set the discretized gamma rates */
{
arb_t s;
arb_init(s);
arb_set_d(s, x->gamma_shape);
if (mode == RATE_MIXTURE_GAMMA)
{
gamma_rates(
rate_mix_rates, x->gamma_categories, s, prec);
}
else if (mode == RATE_MIXTURE_GAMMA_MEDIAN)
{
normalized_median_gamma_rates(
rate_mix_rates, x->gamma_categories, s, prec);
}
else
{
abort(); /* assert */
}
arb_clear(s);
}
/* normalize the rates to account for a possible invariable category */
for (i = 0; i < x->gamma_categories; i++)
{
r = rate_mix_rates + i;
arb_div(r, r, q, prec);
}
arb_clear(p);
arb_clear(q);
}
