The plots allow to check the normal distribution of multiple univaritate marginals by normal quantile-quantile plots. For the different interpretations of amount data a different type of normality is assumed and checked. When an alpha-level is given the marginal displayed in each panel is checked for normality.
# S3 method for acomp
qqnorm(y,fak=NULL,...,panel=vp.qqnorm,alpha=NULL)
# S3 method for rcomp
qqnorm(y,fak=NULL,...,panel=vp.qqnorm,alpha=NULL)
# S3 method for aplus
qqnorm(y,fak=NULL,...,panel=vp.qqnorm,alpha=NULL)
# S3 method for rplus
qqnorm(y,fak=NULL,...,panel=vp.qqnorm,alpha=NULL)
vp.qqnorm(x,y,...,alpha=NULL)
a dataset
a factor to split the dataset, not yet implemented in aplus and rplus
the panel function to be used or a list of multiple panel functions
the alpha level of a test for normality to be performed for each of the displayed marginals. The levels are adjusted for multiple testing with a Bonferroni-correction (i.e. dividing each of the alpha-level by the number of test performed)
further graphical parameters
used by pairs only. Internal use
K.Gerald v.d. Boogaart http://www.stat.boogaart.de
qqnorm.rplus
and qqnorm.rcomp
display qqnorm plots of
individual amounts (on the diagonal), of pairwise differences of amounts
(above the diagonal) and of pairwise sums of amounts (below the
diagonal).
qqnorm.aplus
displays qqnorm-plots of
individual log-amounts (on the diagonal), of pairwise log-ratios of
amounts (above the diagonal) and of pairwise sums of log amount (below the
diagonal).
qqnorm.acomp
displays qqnorm-plots of pairwise log-ratios of
amounts in all of diagonal panels. Nothing is displayed on the
diagonal.
In all cases a joint normality of the original data in the selected
framework would imply normality in all displayed marginal
distributions (although the reciprocal is in general not true!).
The marginal normality can be checked in each of the plots using a
shapiro.test
, by specifying an alpha level. The
alpha level is corrected for multiple testing. Plots displaying a
marginal distribution significantly deviating from a normal
distribution at that alpha level are marked by a red exclamation mark.
vp.qqnorm
is internally used as a panel function to make high dimensional
plots.
plot.acomp
, boxplot.acomp
,
rnorm.acomp
, rnorm.rcomp
,
rnorm.aplus
, rnorm.rplus
data(SimulatedAmounts)
qqnorm(acomp(sa.lognormals),alpha=0.05)
qqnorm(rcomp(sa.lognormals),alpha=0.05)
qqnorm(aplus(sa.lognormals),alpha=0.05)
qqnorm(rplus(sa.lognormals),alpha=0.05)
Run the code above in your browser using DataLab