In order to identify risk factors (e.g. physical inactivity, dietary composition) associated with blood pressure within a given population, it is necessary to adjust for differences in known associations (e.g. age, body weight) using a method such as the analysis of covariance. However, the blood pressure results from the Allied Dunbar National Fitness Survey (ADNFS) were found to be non-linear with age, positively skewed (with heteroscedastic errors) and therefore non-normally distributed. Hence, before valid inferences can be drawn from such data, there is a clear need to formulate an appropriate model for blood pressure that will overcome these undesirable characteristics. A multiplicative model (with allometric body size components) was proposed and fitted to the ADNFS blood pressure results. After a logarithmic transformation the parsimonious solution was able to confirm the association with BMI, the non-linear changes with age, and overcome the heteroscedastic and positively skewed errors, i.e. the residuals from the fitted log-linear models for both systolic and diastolic blood pressure were symmetric and normally distributed. Other factors were found to make a significant additional contribution to the prediction of blood pressure. Cyclists, participants in vigorous physical activity and those subjects who consumed more fresh fruit, rice or pasta, and wine were found to have significantly lower mean levels of blood pressure. Indeed, the gap in blood pressure between participants and non-participants in vigorous physical activity increased further with age. However, subjects who drank more beer tended to have significantly higher mean levels of blood pressure. Thus, by developing an appropriate model for arterial blood pressure, some well known, and some less well known, associations with arterial blood pressure have been identified. The results suggest that physical activity and other lifestyle factors may protect against hypertension.