ESTIMATION IN PARAMETER-REDUNDANT MODELS

The likelihood surface resulting from a parameter-redundant stochastic model is maximised along a completely flat ridge. This ridge may be o rthogonal to some parameter axes, so that these parameters have unique maximum likelihood estimates. For exponential-family models, we show how to determine which parameter combinations are estimable. The appro ach requires the calculation of a derivative matrix and the determinat ion of its null space, both of which are readily achieved in computer algebra packages. Illustrative examples are drawn from the areas of co mpartment modelling and ring-recovery analysis.