TABLE OF CONTENTS Up to this point, the data analysis techniques for linear regression problems were concerned with parameter estimation for an assumed model structure in the regression equation (5.4). The model structure refers to the number and form of the model terms in the regression equation. When the model structure is assumed known and fixed, only the constant model parameter values and their standard errors need to be estimated from the measured data. This can be done using the parameter estimation techniques discussed earlier.
Assuming a model structure also implicitly assumes that each of the terms in the model makes a significant contribution to modeling the variation in the measured response z. Choice of an adequate model structure depends heavily on how the experiment was conducted. However, in many practical cases it is not clear exactly what the model structure should be. From theory, previous experiments, or knowledge of the physical system to be modeled, an analyst can specify candidate regressors that might be considered for the model. Then the task is to select a subset of the candidate regressors that best model the response variable based on the measured data. This procedure is called model structure determination.
In aircraft applications of system identification, the estimation of stability and control derivatives has become a standard procedure for small perturbation flight-test maneuvers where the aerodynamics can be described using regressors that are linear in the independent variables. Interest in near- and post-stall flight regimes, and in dynamics of...
