The structure of Covariance Matrix from Fit Results

[spyhunter01]

Hi all,

Is that the layout of the covariance matrix we actually create when we use these lines in fitting? This picture is for the function with 9 paremeters.

TFitResultPtr r_total = h1->Fit(total, “LRS”);
TMatrixD c_total = r_total->GetCovarianceMatrix();
c_total.Print();

If the answer is YES, then why I can’t get the same result when I use parameter errors directly from root to create my own matrix in excel in this manner?

[Wile_E_Coyote]

It seems to me that Cov(X, Y) = sigma_X * sigma_Y * rho_X_Y, so your Excel table is only valid when the correlation coefficient rho_X_Y = 1 (i.e., when “X” and “Y” are perfectly correlated). Obviously, rho_X_X = 1 (always), so your diagonal elements are always correct.