This study supplemented the National Air Emissions Monitoring Study (NAEMS) with one year of comprehensive measurements of odor emission at five swine and four dairy buildings. The measurements included both standard human sensory measurements using dynamic forced-choice olfactometry and chemical analysis of the odorous compounds using gas chromatography-mass spectrometry. In this article, multilinear regressions between odor and gas concentrations (a total of 20 compounds including H2S, NH3, and VOCs) were investigated. Regressions between odor and gas emission rates were also tested. It was found that gas concentrations, rather than emission rates, should be used to develop multilinear regression models. For the dairy sites, H2S, NH3, acetic acid, propanoic acid, 2-methyl propanoic, and pentanoic acids were observed to be the compounds with the most significant effect on sensory odor. For the swine sites, in addition to these gases, higher molecular weight compounds such as phenol, 4-methyl phenol, 4-ethyl phenol, and 1H-indole were also observed to be significant predictors of sensory odor. When all VOCs were excluded from the model, significant correlations between odor and H2S and NH3 concentrations were still observed. Although these coefficients of determination were lower when only H2S and NH3 were used, they can be used to predict odor variability by up to 83% when VOC data are unavailable.