Physical activity is any bodily movement that results in caloric expenditure. One important aspect of physical activity research is the assessment of usual (i.e., long-term average) physical activity in the population, in order to better understand the links between physical activity and health outcomes. Daily or weekly measurements of physical activity taken from a sample of indivuals are prone to measurement errors and nuisance effects, which can lead to biased estimates of usual physical activity parameters. Fortunately, statistical models can be used to account and adjust for these errors in order to give more accurate estimates of usual physical activity parameters. In this dissertation we develop statistical methods for estimating parameters of usual physical activity. In Chapter 1 we outline metrics and instruments used for physical activity assessment, and review current approaches for modeling usual physical activity and usual dietary intake for regularly consumed food components. In Chapter 2 we develop a model for physical activity data from the National Health and Nutrition Examination Survey (NHANES). A linear regression is defined to model objective monitor-based physical activity as a function of self reported physical activity variables and demographic variables. The fitted model is used to estimate mean daily physical activity levels for demographic groups in the population. In Chapter 3 we develop a method for estimating usual daily energy expenditure parameters from data collected using a self-report instrument and an objective monitoring device. Our method is an extension of existing methods that utilize measurement error models. We illustrate our method with preliminary data from the Physical Activity Measurement Survey (PAMS) collected using a SenseWear Pro armband monitor and a 24-hour physical activity recall.