Missing data for a given cohort of students in a longitudinal study occurs for at least two reasons: either the student has moved or otherwise become unavailable for testing, or the cohort was not in the testing range at a given testing time. A developmental sampling for time of testing x cohort x grade research plan of testing is used to illustrate this point. K. W. Schaie's (1965) proposed set of solutions addressing the developmental models of age, period, and cohort is considered. While the model is logically a three-way situation, computational/logical difficulties led to analyzing three two-way analyses, initially avoiding the use of missing cells and subsequently including missing cells. A model for time of testing x cohort x grade (age) research designs is proposed that follows the full three-way model. After a synthetic data set for grade, cohort, and period for standardized scores is constructed, linear models in a regression solution are applied. To analyze the data for cell, grade, cohort, and period, the sets of two-way design are addressed. The issue of the subjects effect is addressed after the two-way layouts are completed. The problem can then be viewed as a: (1) cohort x time of testing design, (2) grade x time of testing design, and (3) grade x cohort design. The issue of partial repeated measures and consideration of only the repeated measures are discussed. Alternative hypotheses are evaluated, including a time of testing x cohort x grade design with partial subject control with weighted and unweighted means hypotheses. Ten data tables illustrate the different designs. A 17-item list of references is included. (RLC)