eff.aovlist
Compute Efficiency Factors for aovlist Model Terms
Description
Computes the relative efficiency of each treatment term in each stratum.
Usage
eff.aovlist(aovlist)
Arguments
aovlist |
an aovlist object
|
Details
In some multistratum anova, treatment factors are not orthogonal to
error strata, and there is partial information for a treatment
in more than one strata.
Typically, the most efficient strata is
the lowest strata in which the treatment occurs.
The relative efficiencies measure the proportion of
the treatment that projects onto each error stratum.
Efficiencies affect the computation of fitted effects,
means and their standard errors (see model.tables).
The computation requires the qr component of the aovlist object.
Value
a matrix of relative efficiencies: the rows correspond to the error
strata, the columns to the treatment terms.
References
Heiberger, R. M. (1989).
Computation for the Analysis of Designed Experiments.
New York: Wiley.
See Also
Examples
df <- data.frame(
A = factor(c("A1", "A2", "A2", "A1", "A2", "A1",
"A1", "A2"), levels = c("A1", "A2")),
B = factor(c("B2", "B1", "B2", "B1", "B1", "B2",
"B1", "B2"), levels = c("B1", "B2")),
C = factor(c("C1", "C2", "C1", "C2", "C1", "C2",
"C1", "C2"), levels = c("C1", "C2")),
D = factor(c("D1", "D1", "D2", "D2", "D1", "D1",
"D2", "D2"), levels = c("D1", "D2")),
Response = c(0, 0.639, 0.891, 0.983, 0.999, 0.976,
0.93, 0.873))
aovFit <- aov(data=df, subset=-3, Response ~ A + A:B + Error(C/D), qr = TRUE)
eff.aovlist(aovFit)