GPS occultations provide a method of globally profiling water vapor through the lower and mid-troposphere in both clear and cloudy conditions. The profiles represent the water vapor horizontally averaged along the signal path with vertical resolutions of 0.1 to 1 km, precisions of ~0.2 g/kg, and accuracies better than 0.1 g/kg. With the GPS occultation data from the recent CHAMP and SAC-C missions we are able to examine the global distribution and variability of vertically resolved water vapor over the entire annual cycle. We have characterized the observed variability in terms of moments and the full distribution of behavior as inferred from histograms and compared the GPS results with the global Numerical Weather Prediction (NWP) analyses of atmospheric water from NCEP and ECMWF including their respective full distributions. The recent GPS results continue to be drier in general than the analyzed fields in the lower half of the troposphere. The zonal variability estimates are similar but the GPS observed variability tends to be somewhat larger than the analyses. By decomposing the moisture estimates into zonal mean and perturbation terms, we find an anti-correlation exists in the analyses between the true variability and errors suggesting that the lower variability in the analyses is associated with vertical smoothing in the analyses. This is not surprising in that water is known to vary appreciably at vertical scales smaller than can be represented in present global analyses. These results imply the analyses are missing significant variations in the vertical dimension which further suggests important mechanisms may be misrepresented in present analyses and models. Certainly there are indications in our comparisons that suggest vertical diffusion of moisture is too strong in the models.