plot.unidiff: Plot Layer Coefficients From a UNIDIFF Model
In logmult: Log-Multiplicative Models, Including Association Models

plot.unidiff

R Documentation

Plot Layer Coefficients From a UNIDIFF Model

Description

Plots the layer coefficient estimates from a UNIDIFF model, together with confidence bars based on quasi-standard errors or “traditional” standard errors.

Usage

## S3 method for class 'unidiff'
plot(x, what = c("layer.coef", "phi", "maor"),
     se.type = c("quasi.se", "se"),
     conf.int = 0.95, numeric.auto = TRUE, type = "p",
     xlab = names(dimnames(x$data))[3], ylab = NULL,
     add = FALSE, ylim, ...)

Arguments

`x`	an object resulting from a call to `unidiff`
`what`	“layer.coefficient” to plot the layer coefficients in the log odds ratio scale, with a reference of 1 for the first layer; “phi” to plot the intrinsic association coefficient (on the log odds ratio scale); “maor” to plot the mean absolute odds ratio (see `maor`).
`se.type`	whether to use quasi-standard errors or “traditional” standard errors to compute confidence intervals.
`conf.int`	the confidence level to retain for confidence bars.
`numeric.auto`	whether layer names should be converted to numeric values when possible (see “Details” below).
`type`	what type of plot should be drawn: see `plot`. Set to “o” or “b” join points with lines.
`xlab`	a title for the x axis: see see `title`.
`ylab`	a title for the y axis: see see `title`; if `NULL`, an appropriate default is used.
`add`	whether to create a new plot using `plot`, or draw over the existing plot by calling `points` and `segments` directly.
`ylim`	the y limits of the plot.
`...`	Further arguments passed to `plot`.

Details

If numeric.auto = TRUE and layer names (issued from the dimnames of the third dimension of the original table) can be converted to numeric (i.e. they consist of figures), the position of points on the x axis will be determined by the value of the name. This makes most sense when layers represent years, especially when they are not regularly spaced. If this behaviour is disabled, layers will be placed regularly on the x axis, disregarding their possible interpretation as numeric values.