The Effect of Aggregation in Nonlinear Regression.
Technical summary rept.,
WISCONSIN UNIV-MADISON MATHEMATICS RESEARCH CENTER
Pagination or Media Count:
It is often convenient to apply a nonlinear model developed for a single unit to data representing the average response for several units. The relationship between the parameters of the aggregate model and those of the individual units is investigated in a geometrical framework, for a general nonlinear model. The aggregation effect derived is closely related to the bias in nonlinear estimation, as given by Box. Unlike the bias, however, the aggregation effect may be of comparable magnitude to the standard error of the estimates. The theoretical results derived are verified empirically in an application to a model of residential energy consumption. Author
- Economics and Cost Analysis
- Statistics and Probability