Abstract With a network of ground-based ionosondes distributed around the world, the ionospheric peak electron density and its height measured by FORMOSAT-3/COSMIC satellites in terms of GPS radio occultation technique are extensively examined in this article. It is found that, in spite of the latitude, the mean values of the peak electron density measured by COSMIC satellites are systematically smaller than those observed by ground-based ionosondes. The discrepancy between them is dependent on the latitude, namely, it is small in low and mid-latitudes and large in high-latitude region. Moreover, statistical analysis shows that the slopes of the regression line that is best fitted to the scatter diagram of occultation-retrieved peak electron density (ordinate axis) versus ionosonde-observed peak density (abscissa axis) are universally less than one. This feature is believed to be the result of path average effect of non-uniform distribution of the electron density along the GSP ray during the occultation. A comparison between COSMIC-measured peak height and ionosonde-derived peak height hmF2 indicates that the former is systematically higher than the latter. The difference in the two can be as large as 20% or more in equatorial and low-latitude regions. This result implies that the peak height hmF2 derived from the virtual height through true height analysis based on Titheridge method seems to underestimate the true peak height. The correlation between COSMIC and ionosonde peak electron densities is analyzed and the result reveals that correlation coefficient seems to be dependent on the fluctuation of the occultation-retrieved electron density profile. The correlation will be higher (lower) for the electron density profiles with smaller (larger) fluctuations. This feature suggests that the inhomogeneous distribution of the electron density along the GPS ray path during the occultation plays an important role affecting the correlation between COSMIC and ionosonde measurements.