Theoretical analysis of the MIMO outage capacity distribution for an uncorrelated Rayleigh fading channel reveals that the Gaussian distribution is a good approximation, especially for large a number of antennas. The question as to whether the measured channel capacity distribution is close to the theoretical one has not been answered yet in a satisfactory way. We address this problem by developing a statistically-rigorous procedure (in terms of hypothesis testing) for the analysis of the mean and outage capacities of MIMO channels (both, theoretical models and measured) with the main goal being to compare the theoretical and measured capacity distribution, especially the validity of Gaussian approximation for the measured outage capacity. As we demonstrate, there is a tight lower bound on the amount of measured data necessary to provide a unique answer with high confidence probability. Additionally, based on the procedure above, we develop guidelines for future measurements and Monte-Carlo simulations in terms of accuracy.