You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

308 lines
12 KiB
Plaintext

7 months ago
.-
help for ^hplot^
.-
Horizontally labelled plots
---------------------------
^hplot^ varlist [^if^ exp] [^in^ range] [ ^,^
^a^xtol^(^#^) c^start^(^#^) gapm^ag^(^#^) gap^s^(^string^)^
^g^rid ^li^ne ^pts^ize ^r^ange ^so^rt^(^string^)^
^s^ymbol^(^string^)^
^bo^rder ^f^ormat^(^format^) lap nox^axis ^noy^axis ^tti^ck
^xla^bel^(^string^) xli^ne^(^string^) xsc^ale^(^string^)^
^xti^ck^(^string^)^
^b^lank ^gl^egend^(^string^) gllj glpos(^#^) l^egend^(^legendvar^)^
^flipt nit2 t1m(^#^) t1^title^(^string^) t2m(^#^)^
^t2^title^(^string^) tim(^#^) ti^tle^(^string^)^
^fontr(^#^) fontc(^#^) fontrb(^#^) fontcb(^#^)^
^pe^n^(^string^) pent^ext^(^#^) sa^ving^(^graph_filename^)^ ]
Description
-----------
The basic form of ^hplot^ is a graph with one horizontal line for each
observation included. On that line are one or more point symbols
representing the values in varlist according to a common scale.
^hplot^ can produce a variety of horizontally labelled plots for data,
including W.S. Cleveland's dot charts or dot plots; variations on them
with continuous rather than dotted lines; D.R. McNeil's horizontal
parallel line plots; and displays for showing key quantities with or
without confidence intervals.
By default, the data in varlist are represented on horizontal dotted
lines with base at zero that extend to the maximum for each observation.
If negative values are present, dotted lines also extend to the minimum.
Point symbols are used to show actual values.
If the ^grid^ option is used, the data are represented on horizontal
dotted lines that extend over the whole data region of the graph.
If the ^line^ option is used, the data are represented on horizontal
continuous lines with base at the left-hand margin. This can be a lot
faster and may be adequate for exploratory analyses.
If the ^range^ option is used, the data are represented on horizontal
dotted or continuous lines which extend only over the range of values
for each observation, from the smallest value to the largest value.
If the data allow it, a different base may be forced using the ^xscale^
option.
A legend on the left of the data region can be from a specified
variable. If that variable is not specified, the order in the data will
be used; or, if that is not desired, the legend can be blank. The legend
is right-justified. The legend should look readable up to about 30
observations.
For understanding placement of material on the plot, it helps to know
that @gph@ draws in a space defined by 23063 rows and 32000 columns
with origin at top left. See help for @gph@.
Options
--------
Options controlling size, layout and symbol presentation:
---------------------------------------------------------
^axtol(^#^)^ controls the space between the x axis or the position of
the top border and the nearest data line. Default 600.
^cstart(^#^)^ controls the column (horizontal position) of the start of
the lines.
^gapmag(^#^)^ controls the magnitude of any gaps relative to the spacing
between lines. Default 1.
^gaps(^string^)^ places gaps after lines. ^gaps(3,6)^ places gaps after
the 3rd and 6th lines as they appear on the graph, counting down
from the top. ^0^ means a gap before the first line.
^grid^ places point symbols on horizontal dotted grid or guide lines
that extend all the way across the plot. This was originally
recommended by Cleveland if the base is not 0, and later used by him
in all examples.
^sort(^string^)^ means that observations are to be plotted in the
vertical order determined by application of @gsort@. For example,
^sort(^sortvar^)^ indicates sort in ascending order of sortvar
(highest values on bottom) and ^sort(- ^sortvar^)^ indicates
descending order (highest values on top).
^line^ replaces dotted by continuous horizontal lines from the left-hand
margin to each data point.
^ptsize(^#^)^ controls point symbol size. Default 275.
^range^ restricts the dotted or continuous lines to extend from the
minimum to the maximum of those quantities plotted on each
horizontal line.
^symbol(^#^)^ controls point symbols available. The list is most of the
standard list for @graph@, with some additions.
^.^ dot
^o^ small circle
^O^ large circle
^S^ square
^T^ triangle
^d^ diamond
^p^ plus
^i^ invisible
^|^ vertical bar
^,^ short vertical bar
^-^ short horizontal bar
^>^ arrow pointing right
^<^ arrow pointing left
^x^ small cross
^X^ large cross
^a^ arrow pointing from the value for the previous variable
to the value of the present variable
^symbol^ defaults to ^opSdT^ for up to 5 variables and ^o^ for every
variable for 6 or more variables. With 2 or more variables and a
single symbol specified, that symbol is used for every variable.
That is, ^hplot a b c, sy(|)^ expands to ^hplot a b c, sy(|||)^.
Options controlling axes, lines, labels, ticks, border:
-------------------------------------------------------
^border^ adds a border.
^format(^format^)^ controls the format with which ^xlabel^s are shown.
Default %1.0f.
^lap^ (^l^abels ^a^ll ^p^ositive) makes the labels as shown all
positive. ^xla(-40,-20,0,20,40) lap^ will place the labels
40, 20, 0, 20, 40 at the axis positions for -40, -20, 0, 20, 40.
^noxaxis^ suppresses the x axis.
^noyaxis^ suppresses the y axis.
^ttick^ produces short unlabelled ticks on the border above the x axis
that echo the labelled and unlabelled ticks on the x axis. Note that
(unlike the option in @graph@) ^ttick^ with a string is illegal, and
that ^ttick^ necessarily implies ^border^, but not conversely.
^xlabel(^string^)^ controls the labelled ticks on the x axis. Note that
(unlike the option in @graph@) ^xlabel^ without a string is illegal.
numlists may be used, such as ^1/5^ for ^1,2,3,4,5^ and ^0(10)50^
for ^0,10,20,30,40,50^.
^xline(^string^)^ specifies lines drawn for constant values of x. Note
that (unlike the option in @graph@) ^xline^ without a string is
illegal. numlists may be used, such as ^1/5^ for ^1,2,3,4,5^ and
^0(10)50^ for ^0,10,20,30,40,50^.
^xscale(^string^)^ controls the scale of the graph, except that (like
the option in @graph@) it will not cause values to be omitted (for
which purpose use ^if^).
^xtick(^string^)^ controls the unlabelled ticks on the x axis. Note that
(unlike the option in @graph@) ^xtick^ without a string is illegal.
numlists may be used, such as ^1/5^ for ^1,2,3,4,5^ and ^0(10)50^
for ^0,10,20,30,40,50^.
Options controlling legends to left and between gaps:
-----------------------------------------------------
^blank^ blanks out any legend on the left of the data region.
^glegend(^string^)^ places right-justified legend in the gaps between
lines. ^gaps(0,4) glegend(Males!Females)^ places ^Males^ in the gap
before line 1 and ^Females^ in the gap after line 4. Note that ^!^
must be used to separate legends, which thus enables the use of
commas within the legend, but has the side-effect of disallowing the
use of exclamation marks. ^.^ has the special meaning of blank.
^gllj^ makes ^glegend^ left-justified.
^glpos(^#^)^ controls the horizontal position of ^glegend^ and defaults
to an alignment with the main legend.
^legend(^legendvar^)^ specifies a variable to be used for the legend. If
legendvar is a numeric variable with labels, the labels will be used
in the legend.
Options controlling titles (^t1title^, ^t2title^, ^title^):
-----------------------------------------------------
^flipt^ flips titles: ^title^ will be shown at its (default larger) font
size and left-justified at the top of the graph, and ^t1title^ will
be shown at default font size and centred below the axis at the
bottom of the graph (but closer to the axis than the default).
^nit2^: see ^t2title^ below.
^t1title(^string^)^ controls the ^t1title^, shown at default font size
and left-justified at the top of the graph. But see ^flipt^ above.
^t2m(^#^)^ moves the ^t2title^ bodily # to the right. The default is to
start hard left to allow plenty of space for several variable labels
or names in a key, but that default may seem too far left.
^t2title(^string^)^ controls the ^t2title^, shown at default font size
and left-justified at the top of the graph. This defaults to a key
of point symbols if the number of variables is more than 2: the key,
however, is likely to be a mess if the number is more than 5. The
key uses variable labels, or variable names if either they do not
exist or the further option ^nit2^ is invoked. As with @graph@,
^" "^ blanks out the title.
^tim(^#^)^ moves the ^title^ bodily # to the right. The default is to
centre at whatever column would bisect the x axis.
^title(^string^)^ controls the ^title^, shown at its (default larger)
font size and centred below the axis at the bottom of the graph. But
see ^flipt^ above. As with @graph@, ^" "^ blanks out the title.
Other graph options:
--------------------
^fontr(^#^)^ and ^fontc(^#^)^ control the font used for all but the main
title and default to 570 and 290 (which is the default of @gph@).
Font sizes should be changed circumspectly.
^fontrb(^#^)^ and ^fontcb(^#^)^ control the font used for the main
title and default to 923 and 444.
^pen(^string^)^ controls the pens used for data. ^pen^ defaults to ^2^
for every variable. With 2 or more variables and a single pen
specified, that pen is used for every variable. That is,
^hplot a b c, pen(3)^ expands to ^hplot a b c, pen(333)^.
^pentext(^#^)^ controls the pen used for text and other non-data
elements.
^saving(^graph_filename^)^ saves the graph in a .gph file.
Examples
--------
. ^hplot reserves, l(area) xsc(0,30) xla(0(5)30)^
^t1(percent of total) ti(Oil reserves 1994) flipt border^
. ^hplot area, l(name) xsc(0,6) xla(0/6) t1(million^
^square km) ti(Areas of major drainage basins) flipt line^
^fontr(491) fontc(250)^
. ^hplot lcl mean ucl, l(name) xsc(0,6) xla(0/6)^
^t1(95% confidence intervals) sy(|O|) border^
Remarks
-------
In addition to the options above, many choices are coded into ^hplot^
as parameter values. Users may want to copy ^hplot^ and then edit
these permanently or temporarily according to taste.
Cleveland's dot plots are not the same as the histogram-like dotplots
implemented in @dotplot@.
References
----------
Cleveland, W.S. 1984. Graphical methods for data presentation: full
scale breaks, dot charts, and multibased logging. American Statistician
38, 270-80.
Cleveland, W.S. 1994. The elements of graphing data. Hobart Press,
Summit, NJ.
McNeil, D.R. 1992. On graphing paired data. American Statistician
46, 307-11.
McNeil, D.R. 1996. Epidemiological research methods. Wiley, Chichester.
Author
------
Nicholas J. Cox, University of Durham, U.K.
n.j.cox@@durham.ac.uk
Acknowledgments
---------------
Mike Bradburn, Arne Kolstad and Fred Wolfe made very helpful
comments.
Also see
--------
On-line: help for @graph@, @gph@, @format@, @gsort@, @dotplot@,
@hbar@ (if installed), @cihplot@ (if installed),
@tabhplot@ (if installed), @numlist@