eo.iso {UCS} | R Documentation |
Draw an iso-line of a generalised association measure (GAM) in
the (e,o) plane, either for a specified cutoff threshold γ
or an n-best iso-line for a given data set ds
. Optionally,
the corresponding acceptance region can be shaded or filled with solid
colour.
eo.iso(gam, gamma=0, b=1, N=1e6, n.best=NULL, ds=NULL, style=1, fill=solid, solid=FALSE, steps=eo.par("steps"), jitter=eo.par("jitter"), bw=bw, bsearch.min=NULL, bsearch.max=NULL, col=eo.par("col"), lty=eo.par("lty"), lwd=eo.par("lwd"), angle=eo.par("angle"), density=eo.par("density"), solid.col=eo.par("solid"))
gam |
a character string giving the name of a generalised
association measure (GAM). Use the function builtin.gams
from the gam module to obtain a list of available GAMs. |
gamma |
a cutoff threshold that determines the iso-line to be
drawn (by the implicit equation \{g = γ\}). Use the
n.best and ds parameters instead of gamma in
order to obtain an n-best iso-line for the data set ds . |
b, N |
optional balance (b ) and sample size (N )
parameters for GAMs that are not central or size-invariant,
respectively. The default b=1 yields the centralised version
of a non-central GAM (for details, see Evert 2004, Sec. 3.3) |
n.best, ds |
When these parameters are specified, the cutoff
threshold gamma will automatically be determined so as to
yield an n-best acceptance region for the data set
ds . |
jitter |
If TRUE , use jittered coordinates for computing
the n-best cutoff threshold (see above). In this case, the data set
has to be annotated with the add.jitter function first. |
style |
an integer specifying the style (colour, line type and width)
in which iso-lines will be drawn. The number of styles available
depends on the global parameter settings (eo.par ). The
"factory settings" define 5 different styles for iso-lines. |
fill |
If TRUE , fill in the acceptance region bounded by
the given iso-line with shading lines, according to the chosen
style and bw mode . See eo.par for details on
shading styles. |
solid |
If TRUE , fill the acceptance region with solid
colour rather than shading lines, also according to the chosen
style and bw mode. Setting solid=TRUE implies
fill=TRUE . |
steps |
an integer specifying how many equidistant steps are used
for drawing iso-lines. The default value is set with eo.par . |
bw |
If TRUE , the iso-lines are drawn in B/W mode,
otherwise in colour mode. This parameter defaults to the state
specified with the initial eo.setup call, but can be
overridden manually. |
col, lty, lwd |
can be used to override the default style
parameters for iso-lines, which are determined automatically from
the global settings (eo.par ) according to the selected
style and bw mode. |
angle, density |
can be used to override the default style
parameters for shaded acceptance region, which are determined
automatically from the global settings (eo.par ) according to
the selected style and bw mode. |
solid.col |
can be used to override the default colour for
solid filled acceptance regions, which is determined
automatically from the global settings (eo.par ) according to
the selected style and bw mode. |
bsearch.min |
initial lower boundary for binary search algorithm,
when no explicit equation for the iso-line is available
(see gam.iso for details) |
bsearch.max |
initial upper boundary for the binary search algorithm
(see gam.iso ) |
See the eo.setup
help page for a description of the
general procedure used to create (e,o) plots. This help page also has
links to other (e,o) plotting functions. The "factory setting" styles
are described on the eo.par
help page.
The cutoff threshold γ can either be specified explicitly
(with the gamma
parameter) or implicitly as an n-best threshold
(with n.best
, ds
, and optional jitter
). The
latter method produces the same result as
gam.iso(gam, gamma=gamma.nbest(ds, gam, n.best, jitter), ...)
Visualisation by (e,o) iso-lines is most suitable for GAMs that are
both central and size-invariant (see Evert 2004, Sec. 3.3). For
non-central measures, the eo.iso
function uses a balance value
of b = 1, yielding a centralised version of the GAM. Note that
many non-central GAMs (especially those based on statistical tests,
such as log.likelihood
and chi.squared
) have only a weak
dependency on the balance b, so that their centralised
iso-surfaces (i.e. extrusions of the iso-lines along the b-axis) are
very similar to the original iso-surfaces. Other GAMs (most notably
Dice
and similar measures) are highly dependent on b,
though. For measures that are not size-invariant, the sample size is
arbitrarily set to N = 10^6, which is in a realistic range for
real-life data sets. You may wish to modify the default value in
order to match a data set shown in the plot (this is not done
automatically when the ds
parameter is specified), or to
demonstrate the dependency of iso-lines on N.
Evert, Stefan (2004). The Statistics of Word Cooccurrences: Word Pairs and Collocations. PhD Thesis, IMS, University of Stuttgart.
## an example can be found on the "eo.setup" help page