Negative values in covariance matrix

latvietis

Hello

I had measured luminescence decay profile. Then I want to fit a function which would approximate my experimental date. For that I make a simple program in LabWiev. The problem is that, that program give me out a negative values in covariance matrix. Why that?


P.S.
Sorry for my English

maze

Negative covariance is OK. It means that higher-than-average results from one variable will happen at the same time as lower-than-average results from the other variable.

For example, the covariance between how cold it is out and much people get sunburned is probably negative.

If you have more intuition for correlation, this may help: the covariance between 2 variables is just the correlation between the variables, scaled by the standard deviations.

latvietis

Ok

But problem is that the negative values is in diagonal elements. Diagonals elements of covariance matrix is [tex]\sigma^2[/tex]. So I get that errors of functions coefficients [tex]\sigma[/tex] is imaginary.