Summaries in terms of compositions are quite different from classical ones. Instead of analysing each variable individually, we must analyse each pair-wise ratio in a log geometry.
# S3 method for acomp
summary( object, … ,robust=getOption("robust"))a data matrix of compositions, not necessarily closed
not used, only here for generics
A robustness description. See robustnessInCompositions for details. The parameter can be null for avoiding any estimation.
The result is an object of type "summary.acomp"
the mean.acomp composition
a matrix containing the geometric mean of the pairwise ratios
the variation matrix of the dataset ({variation.acomp})
a matrix containing the one-sigma factor for
each ratio, computed as exp(sqrt(variation.acomp(W))). To
obtain a two-sigma-factor, one has to take its squared value (power 1.96, actually).
the inverse of the preceding one, giving the reverse bound. Additionally, it can be "almost" intepreted as a correlation coefficient, with values near one indicating high proportionality between the components.
a matrix containing the minimum of each of the pairwise ratios
a matrix containing the 1-Quartile of each of the pairwise ratios
a matrix containing the median of each of the pairwise ratios
a matrix containing the 3-Quartile of each of the pairwise ratios
a matrix containing the maximum of each of the pairwise ratios
It is quite difficult to summarize a composition in a consistent and interpretable way. We tried to provide such a summary here, based on the idea of the variation matrix.
Aitchison, J. (1986) The Statistical Analysis of Compositional Data Monographs on Statistics and Applied Probability. Chapman & Hall Ltd., London (UK). 416p.
# NOT RUN {
data(SimulatedAmounts)
summary(acomp(sa.lognormals))
# }
Run the code above in your browser using DataLab