Yellow starthistle (Centaurea solstitialis) is a noxious weed that can seriously degrade the quality of grasslands for forage, reduce biodiversity, impede tree seedling recruitment, and change the water budget of ecosystems where it becomes a major component of plant cover. Eustenopus villosus and Chaetorellia succinea were introduced into California as part of a biological control effort to control yellow starthistle. Attack rates of the two insects were measured in 421 locations surveyed in 2001 and 2002. Regression tree analysis and multiple linear regression were performed on the attack rates using a set of climate, land cover, and biological variables. The two methods identified the same variables as the most important predictors of insect attack rates on samples of yellow starthistle taken throughout its range in California. Attack rates of Eustenopus villosus were found to be positively related to precipitation levels, negatively related to the distance from a release site, positively related to elevation, and negatively related to relative humidity. Attack rates of Chaetorellia succinea were found to be negatively related to the attack rates of Eustenopus villosus and precipitation. Regression trees were developed independently of location, but the results of the analysis mapped to geographically contiguous regions. The coefficients of the multiple linear regression models also varied by region. The results show regional predictions for the attack rates of each species given their relationship to the intrinsic and extrinsic variables used.