|
Department of process engineering U218Nonlinear regression - program XREGA |
 |
- improved Marquardt Levenberg algorithms
- multiple nonlinear regression (max. 5 independent variables and arbitrary number of model parameters)
- fully interactive program (matrix editor /column operations, full set of special functions/, user defined models, selection of prepared regression models)
- interactive calculator (transformations of variables etc.)
- enhanced graphic capabilities (a-a, l-a, a-l, l-l axis, multiple functional courses)
Nonlinear parameters estimate (up to 3 nonlinear parameters are selected in a specified domain, the rest of parameters are evaluated by linear regression and resulting sum of squares is displayed as a graph [1 nonlinear parameter], contours [2 parameters] or in a series of figures [3 estimated parameters]) This facility is usefull for location of singularities or multiple local minima of sum of squares. The following data were fitted by the model
Y = B1 + B2 e B3 X 2 + B4 X
and contours in the plane [B3-B4] reveal several of such minimums
@TEC: 3. 3.2003 Change language to
peoples
mailto: Zitny