A comparison of statistical methods for evaluating risk factor changes in community-based studies: An example from the Stanford Three-Community Study