The currently favored method for estimating radii and other parameters of transiting-planet host stars is to match theoretical models to observations of the stellar mean density {rho}, the effective temperature T_eff, and the composition parameter [Z]. This explicitly model-dependent approach is based on readily available observations, and results in small formal errors. Its performance will be central to the reliability of results from ground-based transit surveys such as TrES, HAT, and SuperWASP, as well as to the space-borne missions MOST, CoRoT, and Kepler. Here, I use two calibration samples of stars (eclipsing binaries (EBs) and stars for which asteroseismic analyses are available) having well-determined masses and radii to estimate the accuracy and systematic errors inherent in the {rho}_ method. When matching to the Yonsei-Yale stellar evolution models, I find the most important systematic error results from selection bias favoring rapidly rotating (hence probably magnetically active) stars among the EB sample. If unaccounted for, this bias leads to a mass-dependent underestimate of stellar radii by as much as 4% for stars of 0.4M_{sun}, decreasing to zero for masses above about 1.4M{sun}_. Relative errors in estimated stellar masses are three times larger than those in radii.

Cone search capability for table J/ApJ/709/535/table1 (*Eclipsing binary stars with accurate mass and radius estimates)