The single scanner footprint (SSF) data product produced by the Clouds and the Earth's Radiant Energy System (CERES) instrument on board the Tropical Rainfall Measuring Mission (TRMM) satellite is used to diagnose 68 deep convective cloud objects observed in March 1998. The probability density functions (PDFs) of several observed and retrieved fields from the CERES SSF data product are used to evaluate two sets of simulations produced by a cloud-resolving model. The simulations that use an improved microphysics scheme are found to produce more realistic distributions of albedo, outgoing longwave radiation, cloud top temperature, and cloud height than the control simulations, using an objective measure of the differences between the simulated and observed PDFs. The distribution of cloud optical depth produced by the new simulations is further from that observed than the control simulations. Using a bootstrapping technique, these differences are found to be statistically significant. The reflectivities and precipitation rates associated with the cloud objects observed with the TRMM precipitation radar and TRMM microwave imager are also used to evaluate the two sets of simulations. Both sets are found to produce similar distributions of the precipitation rate. The simulated radar reflectivities produced by the new simulations are somewhat more realistic than those of the control simulations, although both sets produce reflectivities that are too high at altitudes above the freezing level.