Measuring the distance between two points

["Dayan,Y" <Y.Dayan@lse.ac.uk>]

Hi,

I would like to calculate the distance between two points/cases in a 
multivariate dataset.

It seems to me appropriate to use the Mahalanobis distance. 

One variation of the Mahalanobis distance is:

Md(ij)= (Xi-Xj)'inv(C)(Xi-Xj)

Where inv () is the inverse function. 

My questions are:

Is this a correct use/formula of the distance?

Do I take the square root of this? How can I be sure the distance is positive 
given negative correlations? 

Many Thanks,

Yehuda