Learn R Programming

metRology (version 0.9-28-1)

barplot.mandel.kh: Barplot of Mandel's h or k statistics

Description

barplot.mandel.kh produces a bar plot of Mandel's statistics, suitably grouped and with appropriate indicator lines for unusual values.

Usage

# S3 method for mandel.kh
barplot(height, probs = c(0.95, 0.99), main,
		xlab = attr(height, "grouped.by"), 
		ylab = attr(height, "mandel.type"), separators = TRUE, 
		zero.line = TRUE, ylim, p.adjust = "none", 
		frame.plot = TRUE, ..., 
		col.ind = 1, lty.ind = c(2, 1), lwd.ind = 1, 
		col.sep = "lightgrey", lwd.sep = 1, lty.sep = 1, 
		lwd.zero = 1, col.zero = 1, lty.zero = 1)

Arguments

height

An object of class 'mandel.kh' (the name is for consistency with barplot).

probs

Indicator lines are drawn for these probabilities. Note that probs is interpreted as specifying two-tailed probabilities for Mandel's h and one-sided (upper tail) probabilities for Mandel's k.

main

a main title for the plot. If missing, the default is paste(deparse(substitute(x)), " - Mandel's", attr(x, "mandel.type"), if(attr(x, "mandel.method") == "robust") "(Robust variant)")

xlab

a label for the x axis; defaults to the grouped.by attribute for x.

ylab

a label for the x axis; defaults to the mandel.type attribute for x.

separators

Logical; if TRUE, separator lines are drawn between groups of values.

zero.line

logical; if TRUE a horizontal line is drawn at zero.

ylim

the y limits of the plot. For Mandel's k, the default lower limit is zero.

p.adjust

Correction method for probabilities. If not "none", passed to p.adjust prior to calculating indicator lines. Usually, indicator lines are drawn without correction (that is, with p.adjust="none"); specifying a p-value correction effectively turns the Mandel's statistics into single outlier tests.

frame.plot

Logical; If TRUE a box is drawn around the plot.

Other (usually graphical) parameters passed to barplot. Note that some parameters appear after … to prevent spurious argument matching inside barplot.default.

col.ind, lty.ind, lwd.ind

Graphical parameters used for the indicator lines, recyckled to length(probs). For attr(x, "mandel.type")=="h" the graphical parameters are applied to negative as well as positive indicator lines, applied outwards from zero.

col.sep, lwd.sep, lty.sep

Graphical parameters used for the separator lines.

lwd.zero, col.zero, lty.zero

Graphical parameters used for the zero line.

Value

barplot.mandel.kh returns a numeric vector of mid-points of the groups along the x-axis.

Details

Mandel's statistics are traditionally plotted for inter-laboratory study data, grouped by laboratory, to give a rapid graphical view of laboratory bias and relative precision. This plot produces a grouped, side-by-side bar plot.

For classical Mandel statistics, indicator lines are drawn based on qmandelh or qmandelk as appropriate. For robust variants, indicator lines use qnorm for the \(h\) statistic and qf(probs, n, Inf) for the \(k\) statistic. Note that this corresponds to taking the robust estimates of location and scale as true values, so will be somewhat anticonservative.

References

Accuracy (trueness and precision) of measurement methods and results -- Part 2: Basic method for the determination of repeatability and reproducibility of a standard measurement method. ISO, Geneva (1994).

See Also

mandel.h, mandel.k, mandel.kh, pmandelh, pmandelk for probabilities, quantiles etc.

See plot.mandel.kh for the 'classic' Mandel plot.

Examples

Run this code
# NOT RUN {
   data(RMstudy)

   h <- with(RMstudy, mandel.h(RMstudy[2:9], g=Lab))
   barplot(h, las=2) # Lab 4 shows consistent low bias; 
                  # Lab 23 several extreme values.

   #Use colours to identify particular measurands:
   barplot(h, las=2, col=1:8)
   legend("bottomleft", legend=names(h), fill=1:8, cex=0.7, bg="white")
   
   #Example of Mandel's k:
   k <- with(RMstudy, mandel.k(RMstudy[2:9], g=Lab))
   barplot(k, las=2) # Lab 8 looks unusually variable; 
                  # Lab 14 unusually precise
# }

Run the code above in your browser using DataLab