Learn R Programming

metaSEM (version 2.0)

plot: Plot methods for various objects

Description

It plots the models from either the lavaan model or meta, wls, osmasem, and sem (mxsem) objects.

Usage

# S3 method for meta
plot(x, effect.sizes, add.margin = 0.1, interval = 0.95,
    main = "Effect Sizes and their Confidence Ellipses",
    axis.labels = paste("Effect size ", effect.sizes,
        sep = ""), study.col = "black", study.pch = 19,
    study.min.cex = 0.8, study.weight.plot = FALSE,
    study.ellipse.plot = TRUE, study.ellipse.col = "black",
    study.ellipse.lty = 2, study.ellipse.lwd = 0.5,
    estimate.col = "blue", estimate.pch = 18, estimate.cex = 2,
    estimate.ellipse.plot = TRUE, estimate.ellipse.col = "red",
    estimate.ellipse.lty = 1, estimate.ellipse.lwd = 2,
    randeff.ellipse.plot = TRUE, randeff.ellipse.col = "green",
    randeff.ellipse.lty = 1, randeff.ellipse.lwd = 2,
    univariate.plot = TRUE, univariate.lines.col = "gray",
    univariate.lines.lty = 3, univariate.lines.lwd = 1,
    univariate.polygon.width = 0.02, univariate.polygon.col = "red",
    univariate.arrows.col = "green", univariate.arrows.lwd = 2,
    diag.panel = FALSE, xlim = NULL, ylim = NULL, ...)
# S3 method for character
plot(x, fixed.x = FALSE, nCharNodes = 0, nCharEdges = 0,
    layout = c("tree", "circle", "spring", "tree2",
        "circle2"), sizeMan = 8, sizeLat = 8, edge.label.cex = 1.3,
    color = "white", ...)
# S3 method for wls
plot(x, manNames = NULL, latNames = NULL, labels = c("labels",
    "RAM"), what = "est", nCharNodes = 0, nCharEdges = 0,
    layout = c("tree", "circle", "spring", "tree2",
        "circle2"), sizeMan = 8, sizeLat = 8, edge.label.cex = 1.3,
    color = "white", weighted = FALSE, ...)
# S3 method for osmasem
plot(x, manNames = NULL, latNames = NULL, labels = c("labels",
    "RAM"), what = "est", nCharNodes = 0, nCharEdges = 0,
    layout = c("tree", "circle", "spring", "tree2",
        "circle2"), sizeMan = 8, sizeLat = 8, edge.label.cex = 1.3,
    color = "white", weighted = FALSE, ...)
# S3 method for osmasem2
plot(x, manNames = NULL, latNames = NULL, labels = c("labels",
    "RAM"), what = "est", nCharNodes = 0, nCharEdges = 0,
    layout = c("tree", "circle", "spring", "tree2",
        "circle2"), sizeMan = 8, sizeLat = 8, edge.label.cex = 1.3,
    color = "white", weighted = FALSE, ...)
# S3 method for mxsem
plot(x, manNames = NULL, latNames = NULL, labels = c("labels",
    "RAM"), what = "est", nCharNodes = 0, nCharEdges = 0,
    layout = c("tree", "circle", "spring", "tree2",
        "circle2"), sizeMan = 8, sizeLat = 8, edge.label.cex = 1.3,
    color = "white", weighted = FALSE, group = NULL,
    level = NULL, combine = missing(group) && missing(level),
    main.line = 2.5, main.adj = NULL, ...)

Arguments

x

An object returned from either a lavaan model class character, osmasem, osmasem3L, wls or meta

effect.sizes

Numeric values indicating which effect sizes to be plotted. At least two effect sizes are required. To plot the effect sizes of \(y_1\) and \(y_2\), one may use effect.sizes=c(1,2). If it is missing, all effect sizes will be plotted in a pairwise way.

add.margin

Value for additional margins on the left and bottom margins.

interval

Interval for the confidence ellipses.

main

Main title of each plot. If there are multiple plots, a vector of character titles may be used.

axis.labels

Labels for the effect sizes.

study.col

The color for individual studies. See col in par.

study.pch

Plotting character of individual studies. See pch in points.

study.min.cex

The minimum value of cex for individual studies. See cex in par.

study.weight.plot

Logical. If TRUE, the plotting size of individual studies (cex) will be proportional to one over the square root of the determinant of the sampling covariance matrix of the study.

study.ellipse.plot

Logical. If TRUE, the confidence ellipses of individual studies are plotted.

study.ellipse.col

The color of the confidence ellipses of individual studies. See col in par.

study.ellipse.lty

The line type of the confidence ellipse of individual studies. See lty in par.

study.ellipse.lwd

The line width of the confidence ellipse of individual studies. See lwd in par.

estimate.col

The color of the estimated effect size. See col in par.

estimate.pch

Plotting character of the estimated effect sizes. See pch in points.

estimate.cex

The amount of plotting of the estimated effect sizes. See cex in par.

estimate.ellipse.plot

Logical. If TRUE, the confidence ellipse of the estimated effect sizes will be plotted.

estimate.ellipse.col

The color of the confidence ellipse of the estimated effect sizes. See col in par.

estimate.ellipse.lty

The line type of the confidence ellipse of the estimated effect sizes. See lty in par.

estimate.ellipse.lwd

The line width of the confidence ellipse of the estimated effect sizes. See lwd in par.

randeff.ellipse.plot

Logical. If TRUE, the confidence ellipses of the random effects will be plotted.

randeff.ellipse.col

Color of the confidence ellipses of the random effects. See col in par.

randeff.ellipse.lty

The line type of the confidence ellipses of the random effects. See lty in par.

randeff.ellipse.lwd

The line width of the confidence ellipses of the random effects. See lwd in par.

univariate.plot

Logical. If TRUE, the estimated univariate effect sizes will be plotted.

univariate.lines.col

The color of the estimated univariate effect sizes. See col in par.

univariate.lines.lty

The line type of the estimated univariate effect sizes. See lty in par.

univariate.lines.lwd

The line width of the estimated univariate effect sizes. See lwd in par.

univariate.polygon.width

The width of the polygon of the estimated univariate effect sizes.

univariate.polygon.col

The color of the polygon of the estimated univariate effect sizes.

univariate.arrows.col

The color of the arrows of the estimated univariate effect sizes.

univariate.arrows.lwd

The line width of the arrows of the estimated univariate effect sizes.

diag.panel

Logical. If TRUE, diagonal panels will be created. They can then be used for forrest plots for univariate meta-analysis.

xlim

NULL or a numeric vector of length 2; if it is NULL, it provides defaults estimated from the data.

ylim

NULL or a numeric vector of length 2; if it is NULL, it provides defaults estimated from the data.

fixed.x

Argument passed to semPlotModel.

manNames

Names of the observed nodes, in the row order of the RAM filter matrix F. If NULL, names from the fitted model are used.

latNames

Names of the latent nodes, in RAM variable order. If NULL, names from the fitted model are used.

labels

For path diagrams, either labels (the default) to use parameter labels, or RAM to use RAM matrix notation.

what

Argument passed to semPaths

nCharNodes

Argument passed to semPaths

nCharEdges

Argument passed to semPaths

layout

Argument passed to semPaths

color

Argument passed to semPaths

sizeMan

Argument passed to semPaths

sizeLat

Argument passed to semPaths

edge.label.cex

Argument passed to semPaths

weighted

Argument passed to semPaths

group

For plot.mxsem on a multiple-group sem fit (see sem): which group to plot. The original data value (e.g. "site-2", before any sanitizing into a legal mxModel name) always takes precedence over the possibly-sanitized submodel name, via fit$group.map; the sanitized submodel name itself (names(fit$mx.fit@submodels)) only resolves to the matching group when it does not also collide with a different group's original value (e.g. original groups "a.b" and "a_b" both sanitize/map to a submodel named "a_b" or derived from it -- passing "a_b" always means the group whose real data value is "a_b", never the submodel merely named that). When originals and submodel names coincide (the common case, e.g. "g1"), either happens to work, but the original value is the one to reach for. Defaults to the first group. In a combined group and two-level fit, it selects the group whose level is plotted. Ignored otherwise.

level

For plot.mxsem on a two-level or combined group and two-level sem fit (see sem): either "within" (default) or "between", selecting which level to plot. In the between-level diagram, the between components of level-1 variables and the cluster-level observed covariates are drawn as observed variables. Ignored otherwise (including when combine=TRUE). semPlot has no notion of the two levels' relational (joinKey) linkage, so only one level's diagram is shown at a time even when combined (below) into one figure.

combine

For plot.mxsem on a multiple-group or two-level sem fit: if TRUE, plot every group (or both the within and between levels) as side-by-side panels of one figure, overriding group/level. For a combined group and two-level fit, each group is a row with its within and between panels. Defaults to TRUE when neither group nor level is supplied, and to FALSE (one diagram, as selected by group/level) as soon as either of them is -- so plot(fit) shows every panel by default, but plot(fit, group="g2") or plot(fit, level="between") still shows just the one diagram requested without needing combine=FALSE as well. Passing combine=TRUE explicitly still shows every panel regardless of group/level. Ignored otherwise.

main.line

For plot.mxsem on a multiple-group or two-level sem fit: the line argument passed to title for each panel's "within"/ "between" or group-name title. Defaults to 2.5 (pushed up and out of the way of a variance-edge label that semPaths often draws close to the top margin). Ignored otherwise.

main.adj

For plot.mxsem on a multiple-group or two-level sem fit: the adj argument passed to title for each panel's title, controlling its horizontal position (0 = left-aligned, 0.5 = centered, 1 = right-aligned). Defaults to NULL, i.e. title's own default (centered). Since the title is drawn over the diagram rather than reserving its own space, it can still land on top of an edge label that happens to sit centered near the top of that particular model's diagram; nudge it clear with e.g. main.adj=0.8 (or 1 for fully right-aligned) if so. Ignored otherwise.

...

Further arguments passed to the methods.

Author

Mike W.-L. Cheung <[email protected]>

Details

For mxsem objects, a path set by a definition variable (data.*), either directly or through a constraint substituted with replace.constraints=TRUE, is shown at the mean of the variable in the level-1 data of the group plotted. It is not shown for summary input, which has no definition variables. In estimate and standardized diagrams, exactly zero paths use a tiny nonzero graphical weight so they remain visible; the fitted model's estimates are unchanged. The filter matrix F must select distinct observed variables for these plots.

For osmasem and osmasem2 objects, matrices and implied covariances are evaluated at the means of definition variables in the fitted data, after any row selection. For older osmasem2 objects without a saved RAM specification, matrix names are recovered from Fmatrix.

References

Cheung, M. W.-L. (2013). Multivariate meta-analysis as structural equation models. Structural Equation Modeling, 20, 429-454.

See Also

Berkey98, wvs94a meta2semPlot semPaths

Examples

Run this code

# \donttest{
## lavaan model
model <- "y ~ m + x
          m ~ x"
plot(model)
# }

Run the code above in your browser using DataLab