This test is adapted from the Newman-Keuls method. Duncan's test does not control family wise error rate at the specified alpha level. It has more power than the other post tests, but only because it doesn't control the error rate properly. The Experimentwise Error Rate at: 1-(1-alpha)^(a-1); where "a" is the number of means and is the Per-Comparison Error Rate. Duncan's procedure is only very slightly more conservative than LSD. The level by alpha default is 0.05.
model(aov or lm) or answer of the experimental unit
trt
Constant( only y=model) or vector treatment applied to each experimental unit
DFerror
Degree free
MSerror
Mean Square Error
alpha
Significant level
group
TRUE or FALSE
main
Title
console
logical, print output
Details
It is necessary first makes a analysis of variance.
Value
y
class (aov or lm) or vector numeric
trt
constant (only y=model) or vector alfanumeric
DFerror
Numeric
MSerror
Numeric
alpha
Numeric
group
Logic
main
Text
Author(s)
Felipe de Mendiburu
References
1. Principles and procedures of statistics a biometrical approach
Steel & Torry & Dickey. Third Edition 1997
2. Multiple comparisons theory and methods. Departament of statistics
the Ohio State University. USA, 1996. Jason C. Hsu. Chapman Hall/CRC.
See Also
LSD.test, waller.test ,
HSD.test , SNK.test
Examples
library(agricolae)
data(sweetpotato)
model<-aov(yield~virus,data=sweetpotato)
comparison <- duncan.test(model,"virus",
main="Yield of sweetpotato. Dealt with different virus")
duncan.test(model,"virus",alpha=0.01,console=TRUE)
# version old duncan.test()
df<-df.residual(model)
MSerror<-deviance(model)/df
comparison <- with(sweetpotato,duncan.test(yield,virus,df,MSerror, group=TRUE))
print(comparison$groups)