Is taking $3\sigma$ limits for bivariate correlated random variables as the confidence limits an under estimation?

34 Views Asked by At

If the two random variables are uncorrelated, $3\sigma$ limit would mean that the random variables lie inside a hyperrectangle with center coordinates being means of random variables. But if two random variables are correlated, is taking a hyperrectangle instead of corresponding hyperellipsoid an under-approximation or altogether not theoretically correct?