Plots (class "Trellis"
from package lattice) of the
random effects from linear mixed effects model, i.e., the result of
ranef(lme(*))
(of class "ranef.lme"
).
# S3 method for ranef.lme
plot(x, form = NULL, omitFixed = TRUE, level = Q,
grid = TRUE, control, xlab, ylab, strip,
...)
a Trellis plot of the estimated random-effects (coefficients) versus covariates, or groups.
an object inheriting from class
"ranef.lme"
, representing the estimated coefficients or
estimated random effects for the lme
object from which it was
produced.
an optional formula specifying the desired type of plot.
If given as a one-sided formula, a dotplot()
of the estimated random effects (coefficients) grouped according to
all combinations of the levels of the factors named in form
is returned.
If given as a two-sided formula (or by default, NULL
),
an xyplot()
Trellis display of the random
effect (coefficient) versus the named covariates is returned. In
NULL
case the row names of the random effects (coefficients)
are used (as covariates).
See also ‘Details:’.
an optional logical value indicating whether
columns with values that are constant across groups should be
omitted. Default is TRUE
.
an optional integer value giving the level of grouping
to be used for x
. Only used when x
is a list
with different components for each grouping level. Defaults to the
highest or innermost level of grouping.
an optional logical value indicating whether a grid should
be added to plot. Only applies to plots associated with two-sided
formulas in form
. Default is TRUE
.
an optional list with control values for the
plot, when form
is given as a two-sided formula. The control
values are referenced by name in the control
list and only
the ones to be modified from the default need to be
specified. Available values include: drawLine
, a logical
value indicating whether a loess
smoother should be added to
the scatter plots and a line connecting the medians should be added
to the boxplots (default is TRUE
); span.loess
, used
as the span
argument in the call to panel.loess
(default is 2/3
); degree.loess
, used as the
degree
argument in the call to panel.loess
(default
is 1
); cex.axis
, the character expansion factor for
the x-axis (default is 0.8
); srt.axis
, the rotation
factor for the x-axis (default is 0
); and mgp.axis
, the
margin parameters for the x-axis (default is c(2, 0.5, 0)
).
axis labels, each with a sensible default.
optional arguments passed to the Trellis dotplot
function.
José Pinheiro and Douglas Bates bates@stat.wisc.edu
If form
is missing, or is given as a one-sided formula, a
Trellis dot-plot (via dotplot()
from pkg
lattice) of the random effects is generated, with a different
panel for each random effect (coefficient). Rows in the dot-plot are
determined by the form
argument (if not missing) or by the row
names of the random effects (coefficients). Single factors (~g
)
or crossed factors (~g1*g2
) are allowed. For a single factor,
its levels determine the dot-plot rows (with possibly multiple dots
per row); otherwise, if form
specifies a crossing of factors, the dot-plot rows are determined by
all combinations of the levels of the individual factors in the
formula.
If form
is a two-sided formula, the left hand side must
be a single random effect (coefficient) and the right hand side is
formed by covariates in x
separated by +
.
An xyplot()
Trellis display is generated,
with a different panel for each variable listed in the right hand side
of form
. Scatter plots are generated for numeric variables and
boxplots are generated for categorical (factor
or
ordered
) variables.
ranef.lme
, lme
,
dotplot
.
fm1 <- lme(distance ~ age, Orthodont, random = ~ age | Subject)
plot(ranef(fm1))
fm1RE <- ranef(fm1, augFrame = TRUE)
plot(fm1RE, form = ~ Sex)
plot(fm1RE, form = age ~ Sex) # "connected" boxplots
Run the code above in your browser using DataLab