[R] Why do I get (not estimable) in this aov data

[Hedberg Peter]

I realize the problem now, and you are right aov is wrong.
It seams that it was completely wrong to use the transect in a fixed formula, since my Trasects are labeled A to P, and A to D is in one mainplot, E-H in another, I-L an a third, and M-P in a fourth. In this case lme, with random variables for Transect is better. Thank you for the help.

Best regards, Petter



Peter Dalgaard <pdalgd at gmail.com> skrev :

> On Nov 30, 2010, at 13:58 , Hedberg Peter wrote:
> 
> >> aov_data
> > Call:
> >   aov(formula = Species1 ~ Site + Obstacle + Treatment + as.factor(Dist_Obstacle) + 
> >    as.factor(Dist_start) + Transect + Mainplot + Obsplot)
> > 
> > Terms:
> >                     Site  Obstacle Treatment as.factor(Dist_Obstacle) as.factor(Dist_start)  Transect   Obsplot Residuals
> > Sum of Squares   2143.984   446.274   340.042                  736.073               173.707   800.270  4014.378 17238.625
> > Deg. of Freedom         2         1         1                        4                    3        10        60       271
> > 
> > Residual standard error: 7.97566 
> > 27 out of 109 effects not estimable
> > Estimated effects may be unbalanced
> > 
> > 
> > 
> > My question is why do I get "effects not estimable", and "effects may be unbalanced). I have checked the data and it is balanced.
> 
> Unfortunately, your attachment did not contain the data, but the sum of the Deg. of Freedom  above is 352, suggesting that the observation count is 353, which is prime, so I find it difficult to believe that you have balanced data in the sense of a complete factorial design, even for a subset of your factors. A complete factorial with those DF would take more than 160000 observations!
> 
> I suspect that aov() is simply the wrong tool for these data. lm() will do it, but watch out for the aliased effects.
> 
> -- 
> Peter Dalgaard
> Center for Statistics, Copenhagen Business School
> Solbjerg Plads 3, 2000 Frederiksberg, Denmark
> Phone: (+45)38153501
> Email: pd.mes at cbs.dk  Priv: PDalgd at gmail.com

___________________________________
NOCC, http://nocc.sourceforge.net