[R] Coefficient of determination for non-linear equations system (nlsystemfit)

[Arne Henningsen]

Dear Antti

On 20 November 2012 10:24, Antti Simola <asimola78 at gmail.com> wrote:
> I have estimated system of three linear equations with one non-linear
> restrictions with nlsystemfit.

Please read the FAQ at http://www.systemfit.org/

> I was wondering how I can calculate the
> R-squared (or some alternative coefficient of determination) for the whole
> system. This is automatically given by linear systemfit but not by
> nlsystemfit. I can get the values for each of the equations separately, but
> apparently not for the whole system.

You have to do this manually, e.g. by equation (43) in the paper:
   http://www.jstatsoft.org/v23/i04/paper

> I'm also wondering why separate equations' objects all appear in together
> with a command like:
>
> nl.system <- nlsystemfit()
>
> nl.system$eq[i]
>
> but e.g. the following produces NULL as the value for the individual
> objects, e.g. R-squared

As the "eq" component is a "list" (see documentation), you must use
double brackets:
   nl.system$eq[[i]]

> eq.1 <- nl.system$eq[1]
>
> eq.1$r2

You can directly use:
   nl.system$eq[[i]]$r2

Best wishes from Copenhagen,
Arne

--
Arne Henningsen
http://www.arne-henningsen.name