Windows 2000 . 5.00.2195 with Service Pack 1.
R 1.5.1
Output from my split-split plot aov "alerted" me that I have done
something
wrong. I designed an experiment with all combinations of all levels of each
treatment, but lost a little data (3 out of 192 plots). With the following
data, I run the following model:
> collim[c(1:6,187:192),c(1,3:6,9)]
plot Litter Fert Seed S density
X192.4 192 R 16 C 7 12
X192.3 192 C 16 C 14 26
X192.2 192 C 16 A 8 3
X192.1 192 R 16 A 9 12
X191.4 191 R 32 A 3 0
X191.3 191 C 32 A 10 11
.....
X146.2 146 R 32 C 7 7
X146.1 146 C 32 C 7 3
X145.4 145 R 8 C 9 21
X145.3 145 C 8 C 6 15
X145.2 145 R 8 A 9 10
X145.1 145 C 8 A 8 14
fm1 <- (log(S) ~ Fert*Seed*Litter*density+ Error(plot/Seed)
where Fertilizer is applied to entire plots, Seed is applied at two
densities to nested subplots, and dead plant Litter is removed or not from
sub-subplots within each Seed subplot. Variables plot, Fert, Seed, and
Litter are factors, density and S (number of species present) are continuous.
The missing plots: 2 that received 16 g fertilizer, no seeds, and litter
intact, 1 plot received no fertilizer, seeds added, and litter intact.
The output, below, tests All factors over Error: Within. I don't
understand. I expected Fert to be tested over plot residuals, Seed and
Seed:Fert tested over plot:Seed residuals, and all else tested over Within
residuals.
Can any body help?
summary(fmc)
Error: plot
Df Sum Sq Mean Sq
Fert 1 0.98913 0.98913
Error: plot:Seed
Df Sum Sq Mean Sq
Fert 1 0.02016 0.02016
Error: Within
Df Sum Sq Mean Sq F value Pr(>F)
Fert 3 16.0269 5.3423 59.2169 < 2.2e-16 ***
Seed 1 0.0052 0.0052 0.0579 0.81009
Litter 1 0.0834 0.0834 0.9242 0.33787
density 1 5.0877 5.0877 56.3953 4.38e-12 ***
Fert:Seed 3 0.0203 0.0068 0.0750 0.97336
Fert:Litter 3 0.0984 0.0328 0.3637 0.77930
Fert:density 3 0.7831 0.2610 2.8933 0.03719 *
Seed:Litter 1 0.0132 0.0132 0.1468 0.70217
Seed:density 1 0.0343 0.0343 0.3805 0.53826
Litter:density 1 0.2995 0.2995 3.3198 0.07038 .
Fert:Seed:Litter 3 0.0431 0.0144 0.1592 0.92361
Fert:Seed:density 3 0.9942 0.3314 3.6734 0.01359 *
Fert:Litter:density 3 0.3865 0.1288 1.4280 0.23674
Seed:Litter:density 1 0.0040 0.0040 0.0440 0.83410
Fert:Seed:Litter:density 3 0.1477 0.0492 0.5457 0.65175
Residuals 155 13.9834 0.0902
---
Martin Henry H. Stevens, Assistant Professor
338 Pearson Hall
Botany Department
Miami University
Oxford, OH 45056
Office: (513) 529-4206
Lab: (513) 529-4262
FAX: (513) 529-4243
http://www.muohio.edu/~botcwis/bot/henry.html
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
"Martin Henry H. Stevens" <HStevens at muohio.edu> writes:> The output, below, tests All factors over Error: Within. I don't > understand. I expected Fert to be tested over plot residuals, Seed and > Seed:Fert tested over plot:Seed residuals, and all else tested over > Within residuals. > > Can any body help? > > summary(fmc) > > Error: plot > Df Sum Sq Mean Sq > Fert 1 0.98913 0.98913 > > Error: plot:Seed > Df Sum Sq Mean Sq > Fert 1 0.02016 0.020161DF total for these two strata makes me suspect that you put in "plot" as a numerical variable rather than as a factor. -- O__ ---- Peter Dalgaard Blegdamsvej 3 c/ /'_ --- Dept. of Biostatistics 2200 Cph. N (*) \(*) -- University of Copenhagen Denmark Ph: (+45) 35327918 ~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk) FAX: (+45) 35327907 -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- 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 _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
Dear Peter, I did double check whether all factors had "factor" as their class. "plot" is a factor, with 48 levels. I used Professor Ripley's suggestion to go with lme() which can handle unbalanced designs. It had no problem (that I was aware of!). Its format and methods, however, are less familiar to me, and so I was resisting it. Now, as a consequence, however, I am relearning (again!) the relations among variance components, Mean squares, etc. I am still not clear, though, how lme() tests fixed effects. I'll get it yet. Thank you for your feed back. At 04:45 PM 6/19/2002, Peter Dalgaard BSA wrote:>"Martin Henry H. Stevens" <HStevens at muohio.edu> writes: > > > The output, below, tests All factors over Error: Within. I don't > > understand. I expected Fert to be tested over plot residuals, Seed and > > Seed:Fert tested over plot:Seed residuals, and all else tested over > > Within residuals. > > > > Can any body help? > > > > summary(fmc) > > > > Error: plot > > Df Sum Sq Mean Sq > > Fert 1 0.98913 0.98913 > > > > Error: plot:Seed > > Df Sum Sq Mean Sq > > Fert 1 0.02016 0.02016 > >1DF total for these two strata makes me suspect that you put in "plot" >as a numerical variable rather than as a factor. > >-- > O__ ---- Peter Dalgaard Blegdamsvej 3 > c/ /'_ --- Dept. of Biostatistics 2200 Cph. N > (*) \(*) -- University of Copenhagen Denmark Ph: (+45) 35327918 >~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk) FAX: (+45) 35327907Martin Henry H. Stevens, Assistant Professor 338 Pearson Hall Botany Department Miami University Oxford, OH 45056 Office: (513) 529-4206 Lab: (513) 529-4262 FAX: (513) 529-4243 http://www.muohio.edu/~botcwis/bot/henry.html -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- 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 _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._