A comparison of methods to compute the “effective duration” of the autocorrelation function and an alternative proposal