int main(){
std::vector<double> const array1{
1.2, 2.0, 2.5, 4.0, 3.0, 6.0, 5.5, 6.3, 7.1, 5.4
};
std::vector<double> const array2{
3.4, 3.3, 3.0, 5.5, 1.2, 2.4, 3.2, 3.1, 2.9, 3.2
};
{
double const covarianceValue = covariance<double>(
array1,
array2
);
assert(feq(covarianceValue, -0.108889));
}
{
double const covarianceValue = gsl_stats_covariance(
&array1.front(), 1,
&array2.front(), 1, array2.size()
);
assert(feq(covarianceValue, -0.108889));
}
}
CAMLprim value ml_gsl_stats_covariance(value data1, value data2)
{
size_t len = Double_array_length(data1);
double result;
check_array_size(data1, data2);
result = gsl_stats_covariance(Double_array_val(data1), 1,
Double_array_val(data2), 1, len);
return copy_double(result);
}
double correlation(double *X, double *Y, int count)
/* Compute the correlation between X and Y
correlation(X,Y) = cov(X,Y)/ squareRt(var(X)var(Y))
page 332 Sheldon Ross "A First Course in Probability" 1998
*/
{
double varX = gsl_stats_variance(X, 1, count);
double varY = gsl_stats_variance(Y, 1, count);
double covXY = gsl_stats_covariance(X, 1, Y, 1, count);
double correlation = covXY / sqrt(varX *varY);
return correlation;
}
/* Example from R (GPL): http://www.r-project.org/
* (samples <- matrix(c(4.348817, 2.995049, -3.793431, 4.711934, 1.190864, -1.357363), nrow=3, ncol=2))
* cov(samples) # 19.03539 11.91384 \n 11.91384 9.28796
*/
static int
multivar_vcov (const double data[], size_t d, size_t tda, size_t n,
double vcov[], size_t tda2)
{
size_t j1 = 0, j2 = 0;
for (j1 = 0; j1 < d; ++j1)
{
vcov[j1 * tda2 + j1] = gsl_stats_variance(&(data[j1]), tda, n);
for (j2 = j1 + 1; j2 < d; ++j2)
{
vcov[j1 * tda2 + j2] = gsl_stats_covariance(&(data[j1]), tda,
&(data[j2]), tda, n);
vcov[j2 * tda2 + j1] = vcov[j1 * tda2 + j2];
}
}
return GSL_SUCCESS;
}
