Re: Response to Gonzalo

[BILL MITCHELL <ECWFM@xxxxxxxxxxxxxxxxxxx>]

Paul repeatedly makes this point:


>where you went wrong with [precise linear models. The trouble with
>non-linear models is, as I have said several times but you simply ignore,
>is that in any time series of points of observations, a nonlinear line
>can always be found that will "fit" the observations better than a linear
>line.  In other words if I have a scatter diagram, of ten points which,
>for the sake of discussion, are randomly distributed around a straight
>line, I can always find some line with nine turning points that will
>"fit" the data better than the straight line. That nonlinear nine-turning
>point curve is more precise in describing the position of the
>observations -- So what?

it is not correct of-course, although it depends on the meaning of "fit". the
linear model will be a more parsimonious model and therefore use the degrees of
freedom more efficiently, than a model which uses up all the data points (and
face breakdown of estimation).

kind regards
bill
*******************************************************************************
 William F. Mitchell            Telephone: +61-49-215027      .-_|\
 Department of Economics                   +61-49-705133     /     \    about
 The University of Newcastle    Fax:       +61-49-216919     \.--._/*<-- here
 Callaghan   NSW  2308                                            v
 Australia                      Email : ecwfm@xxxxxxxxxxxxxxxxxxx

 World Wide Web Home Page: http://econ-www.newcastle.edu.au/~bill/billyhp.html
*******************************************************************************