Next: break Up: MacAnova Help File Previous: boxcox()   Contents


boxplot(x1,x2,...,xk [,vs:indv, boxsize:W] [,vertical:T, excludeM:T,
  graphics keyword phrases]), x1,...,xk REAL vectors, indv REAL length k
  vector with no MISSING values, W REAL non-negative vector or scalar
boxplot(Struc, [,vs:indv, boxsize:W] [, vertical:T, graphics keyword
  phrases]), Struc a structure with k REAL vector components

Keywords: plotting, descriptive statistics
boxplot(x1, x2, ... , xk) produces horizontal parallel Tukey boxplots
for the vectors x1 through xk and plotting positions 1, 2, ..., k on the
y axis.

boxplot(x1, x2, ... , xk,vertical:T) and boxplot(Struc,vertical:T) do
the same except the boxplots are aligned vertically at plotting
positions 1, 2, ..., k on the x axis.  Pre-defined macro vboxplot()
which is used identically to boxplot(), makes use of the feature to make
vertical boxplots.

boxplot(x1, x2, ..., xk, vs:Predictor [,vertical:T] ...) does the same
except the boxes are aligned with Predictor[1], Predictor[2], ...,
Predictor[k] on the y or x axis.  Predictor must be a REAL vector with
no MISSING values with length(Predictor) = k.

boxplot(x1, x2, ..., xk, boxsize:W, ...) does the same except the
thickness of the boxes is determined by non-negative REAL scalar or
vector W.  See below for details.

boxplot(Struc, ....) produces parallel box plots for the components of
structure Struc, all of which must be REAL vectors.  You can use any
boxplot() or graphical keywords..

boxplot(split(y,a) [,vertical:T] ... ) draws parallel box plots of the
data in vector y classified according to levels of factor a.  See

boxplot(split(y) [,vertical:T], ...) draws parallel box plots of the
data in each column of matrix y.

You can use keyword phrase boxsize:W, where W is a non-negative scalar
or vector to specify the "thickness" of the boxes (height for horizontal
boxes, width for vertical boxes).  When W is not a scalar, length(W)
must match the number of boxes.  When W is a vector and W[j] = 0, box j
is omitted.

When you don't use keyword 'boxsize', the default thickness is such that
about 2/3 of the space between the first and last box is made of up
boxes and 1/3 of interbox space.

When keyword phrase 'excludeM:T' is an argument, and the number of
non-MISSING values in a sample is odd, the median is omitted in
calculating the quartiles as the medians of the upper and lower halves.

You may use keywords 'dumb', 'xmin', 'xmax', 'ymin', 'ymax', 'logx',
'logy', 'xlab', 'ylab', 'title', 'xaxis', 'yaxis', 'borders', 'ticks',
'xticks', 'yticks', 'xticklen', 'yticklen', 'xticklabs', 'yticklabs',
'height', 'width', 'pause', 'silent' and 'notes' as for other plotting
commands.  See topics 'graph_keys', 'graph_border' and 'graph_ticks'

When option 'dumbplot' has been set False (see subtopic
'options:"dumbplot"'), the plot will be a low resolution plot unless
'dumb:F' is an argument.

Note: Using 'logx:T' and/or 'logy:T' affects only the scaling used in
plotting, not the determination of outliers.  With logy:T without
vertical:T or logx:T with vertical:T`q, the thickness of the boxes will
be affected since the edges are equally distant from the middle in
arithmetic units but not in logarithmic units

See topic 'graph_files' for information on how to save a boxplot in a
file using keywords 'file', 'new, 'ps', 'screendump', and 'epsf'.

See topics 'macintosh' and 'wx' for information on how to print graphs
in windowed versions (Macintosh, Windows, Motif).

See also topics showplot(), 'structures', 'graph_keys'.

Gary Oehlert 2003-01-15