ABSTRACT

In longitudinal data analysis, where the goal is to estimate population or individual-specific averages over time, it is often the case that the repeated observations from a single individual can be viewed as arising from sampling a smooth underlying individual mean curve together with non-smooth deviations. The deviations can be thought of as a combined result of measurement error and biologic variability that would fluctuate upon resampling the same individual over the same period of time. When the assumption of smooth individual mean curves is reasonable, it makes sense to incorporate it into the estimation process in order to capitalize on the variance reduction that typically ensues when correct assumptions are used to reduce the effective degrees of freedom used by a statistical model.