Comparing Traditional Longitudinal Modeling Strategies of Forestry with Mixed-Effects Models: Restrictions in Model Formulation
Generalized algebraic difference approach and site index models, which are commonly fitted to repeated-measures data in forestry, are formulated by specifying that one or more regression parameters depend on a single, common subject-specific quantity. Their formulation leads to shortcomings
not present with “effects” types of formulations, which involve specifying each subject-specific regression parameter as depending on at least one local quantity that no other parameters depend on. We give conceptual illustrations as well as empirical examples using published models.