Camera trapping is a powerful tool for studying mammal populations over increasingly large spatial scales. Density estimation is a commonly desired outcome, but most approaches only work for species that can be individually recognized, and researchers are typically constrained to measures of site occupancy or detection rate instead; metrics that are presumed to be directly related to animal density. To test relationships between these metrics we estimated density, occupancy and detection rate of male white-tailed deer (Odocoileus virginianus) using the same set of camera trap data collected from 1,199 unbaited cameras at 20 study sites. We found a strong (R2 = 0.80) positive relationship between detection rate and density and low (R2 = 0.27) correlation between occupancy and density. We identified two sites, both urban parks, where the relationship between camera trap detections and density estimates did not correspond, presumably because of heterogeneous movement patterns. When paired hunted and unhunted parks were compared, estimate agreement was least for pairs with the highest differences in surrounding housing density. Ecological models explaining variation in occupancy and detection rate were similar, but different from density models, suggesting that the metrics are reflecting different ecological relationships. Detection rate performed well as a proxy for density across large scales but poorly at small scales while occupancy showed the opposite trend. Our results indicate that managers should not use detection rate and occupancy as proxies for density when comparing between areas where animals might be expected to move or behave differently, such urban vs. wild landscapes.