We attempt to recover the mean vertical velocity and vertical velocity dispersion as a function of the Galactic height for a sample drawn from a realistic Galaxy distribution function by following the method presented by Moni Bidin, Carraro & Mendez. We find that, for the sample size used, the observational error in the velocities is much smaller than the Poisson noise which has not been accounted for by Moni Bidin et al. We repeat the analysis on a large number of samples to estimate the contribution of the Poisson noise and to uncover any systematics. We find that the dispersion is systematically overestimated at low Galactic heights and slightly underestimated at high Galactic heights, leading to an underestimate of the gradient of the dispersion with Galactic height. The causes of the systematics are revealed by repeating the calculation using a method inspired by Girard et al. This method recovers the expected dispersion much more successfully and in particular yields a gradient of the dispersion with Galactic height which is approximately three times that found using the method presented by Moni Bidin et al.