balance in AoV (was [R] aov() and NaN)
ripley@stats.ox.ac.uk
ripley at stats.ox.ac.uk
Fri Jul 5 13:04:56 CEST 2002
The results are as one should expect: they are not `strange'. I'm afraid
I can't help you, as I do not know what you are failing to understand
about multistratum models (or what you/SPSS are doing differently).
Hint 1: `balance' in the key concept.
Hint 2: in the absence of balance, modern texts recommend a REML or ML
analysis, and lme can do that.
On Fri, 5 Jul 2002, Jan Malte Wiener wrote:
> ripley at stats.ox.ac.uk wrote:
> > aov does not. It's na.action argument does. It is dscribed on the help
> > page for lm, which ?aov refers to.
> >
> > BTW, missing values in R are not the same thing as NaNs, but for this
> > purpose NaNs are treated as missing.
> >
> > On Fri, 5 Jul 2002, Jan Malte Wiener wrote:
> >
> >
> >>hi,
> >>another anova question !!
> >>how does aov() treat missing values ?
> >>
> >>since it might depend on my experimental design:
> >>i have a within subject design with two factors ...
> >>
> >>i use aov() like this:
> >>
> >>aov(performance~type*block+Error(subj/(type+block)),data=anovaData)
> >>
> >>now it might happen that i have a couple of NaN data points in my result
> >>dataFrame.
> >>how can i know/control how aov() is treating those missing values ?
> >>
> >>greetinx jan
>
>
> Yes, I tried using na.action=na.exclude (na.fail, na.omit).
> obviously with na.fail I do not get any results, but with the other
> arguments I always get the same results, which are different from spss
> results
> The results are not only different from spss, aov() behaves strange also
> (look at the "Error: subj:type" or "Error: subj:block" section ...)
>
> summary(aov(perf~type*block+Error(subj/(type+block)),
> na.action=na.omit,data=anovaDataAllSubjCorrNav))
>
>
> Error: subj
> Df Sum Sq Mean Sq F value Pr(>F)
> type 1 0.25032 0.25032 2.4855 0.1314
> Residuals 19 1.91355 0.10071
>
> Error: subj:type
> Df Sum Sq Mean Sq F value Pr(>F)
> type 1 0.03227 0.03227 0.3140 0.5818
> block 1 0.15184 0.15184 1.4774 0.2391
> Residuals 19 1.95268 0.10277
>
> Error: subj:block
> Df Sum Sq Mean Sq F value Pr(>F)
> block 1 0.06253 0.06253 0.8086 0.3798
> type:block 1 0.22826 0.22826 2.9520 0.1020
> Residuals 19 1.46914 0.07732
>
> Error: Within
> Df Sum Sq Mean Sq F value Pr(>F)
> type:block 1 0.00413 0.00413 0.0726 0.7892
> Residuals 34 1.93538 0.05692
>
>
> #####
> for the section: "Error : subj:type" I get results for type and block
> factor ?? That I do not understand and that does not happen when having
> no NaN's -> see below !
> ######
>
>
> > print(summary(aov(perf~type*block+Error(subj/(type+block)),
> na.action=na.omit, data=anovaDataAll)) )
>
> Error: subj
> Df Sum Sq Mean Sq F value Pr(>F)
> Residuals 24 2.9711 0.1238
>
> Error: subj:type
> Df Sum Sq Mean Sq F value Pr(>F)
> type 1 0.00111 0.00111 0.0132 0.9096
> Residuals 24 2.02667 0.08444
>
> Error: subj:block
> Df Sum Sq Mean Sq F value Pr(>F)
> block 1 0.21778 0.21778 4.3596 0.04759 *
> Residuals 24 1.19889 0.04995
> ---
> Signif. codes: 0 `***' 0.001 `**' 0.01 `*' 0.05 `.' 0.1 ` ' 1
>
> Error: Within
> Df Sum Sq Mean Sq F value Pr(>F)
> type:block 1 0.01000 0.01000 0.2424 0.6269
> Residuals 24 0.99000 0.04125
>
>
>
>
> Thanks for comments on this,
> Jan Wiener
>
>
>
> --
> Jan Malte Wiener
> Max-Planck-Institute for Biological Cybernetics
> Spemannstr. 38, 72076 Tuebingen, Germany
> Tel.: +49 7071 601 631
> Email: jan.wiener at tuebingen.mpg.de
>
>
--
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 272860 (secr)
Oxford OX1 3TG, UK Fax: +44 1865 272595
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !) To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
More information about the R-help
mailing list