[R] Residuals in gls

Prof Brian Ripley ripley at stats.ox.ac.uk
Fri Apr 15 20:17:52 CEST 2005


The `correct thing to do' even for lm is to use the generic accessor 
function residuals().   It may give a different answer even for lm 
(depending on na.action).

Similarly, use fitted() for the fitted values.

On Fri, 15 Apr 2005, Panos Hadjinicolaou wrote:

> I am doing a multiple linear regression of an ozone time-series on time and
> other explanatory variables. I have been using the "lm" model but I am recently
> experimenting with "gls".
>
> With the "lm" model I was able to look at the residuals by $res in the "summary
> (lm(...))" and then check with "acf" for autocorrelation in these residuals
> which had the same "length" with the original time-series.
>
> A similar action in "gls" would return only 5 values for residuals
> (Min,Q1,Med,Q3,Max). So I am estimating the residuals for the predicted
> time-series by subtracting $fitted from my original time-series. I would like to
> ask if this a correct thing to do.

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595




More information about the R-help mailing list