Statistics is concerned with guessing the unknown. A challenge in statistics is integrating classical frequentist based statistical guesses with Bayesian methods guesses. This challenge is particularly apparent when using mixed models for longitudinal data analysis, since empirical Bayes estimators such as the best linear unbiased predictor are often recommended. A simple example illustrates the dilemma. A person is asked via a 24 food record to record the total Kcal ingested. The result is a value of 6000 Kcal, well outside the range of usual values. What is the best guess of the person's typical intake? 6000 Kcal, or a value closer to the "population" norm? This broad area is the focus of my research.