[R] Forcing coefficents in lm(), recursive residuals, etc.

[Rick Ram]

Hi all,
Just to clarify, I know that the predict() function would the normal
avenue for applying a model but the problem is that I need to
calculate recursive residuals, and the recresid() function needs an
object with class "lm".
Best,
R.

On 28/07/05, Rick Ram <r.ramyar at gmail.com> wrote:
> Sorry guys, resending this - none of my posts have gone through
> because HTML emails where not being delivered... sending this
> plaintext now!
> 
> On 28/07/05, Rick Ram <r.ramyar at gmail.com> wrote:
> > Resending cos I think this didn't get through for some reason... apologies
> > if it arrives twice!
> >
> >
> > ---------- Forwarded message ----------
> > From: Rick Ram < r.ramyar at gmail.com>
> > Date: 28-Jul-2005 18:03
> > Subject: Forcing coefficents in lm(), recursive residuals, etc.
> > To: R-help <r-help at stat.math.ethz.ch >
> >
> > Hello all,
> >
> > Does anyone know how to constrain/force specific coefficients when running
> > lm()?
> >
> > I need to run recresid() {strucchange package} on the residuals of
> > forecast.lm, but forecast.lm's coefficients must be determined by
> > parameter.estimation.lm
> >
> > I could estimate forecast.lm without lm() and use some other kind of
> > optimisation, but recresid() requires an object with class lm.  recresid()
> > allows you to specify a formula, rather than an lm object, but it looks like
> > coefficients are estimated this way too and can't be forced.
> >
> > Here is a bit of code to compensate for my poor explanation:.
> >
> > # Estimate the coefficients of model
> > parameter.estimation.lm = lm(formula = y ~ x1 + x2, data =
> > estimation.dataset)
> > # How do I force the coefficients in forecast.lm to the coeff estimation
> > from parameter.estimation.lm??
> >
> > forecast.lm = lm(formula = y ~ x1 + x2, data = forecast.dataset)
> > # Because I need recursive residuals from the application of the
> > coefficients from parameter.estimation.lm to a different dataset
> > recresid(forecast.lm)
> >
> > Thanks in advance guys,
> >
> > R.
>