effectplot: Plot for regressions to compare effects sizes of normalized...

In skranz/regtools: Some tools for regressions and presentation of regressions results

Description

The plot shall help to compare magnitudes of the influence of different explanatory variables. The default effect is "10-90", i.e. the effect of when -ceteris paribus- changing an (numeric) explanatory variable from its 10

Usage

effectplot(reg, dat = get.regression.data(reg, source.data = source.data),
  source.data = NULL, main = NULL, vars = intersect(colnames(dat),
  names(coef(reg))), ignore.vars = NULL, numeric.effect = "10-90",
  dummy01 = TRUE, sort = TRUE, scale.depvar = NULL,
  depvar = name.of.depvar(reg),
  xlab = "Explanatory variables\n(low baseline high)",
  ylab = paste0("Effect on ", depvar, ""), colors = c(pos = "#11AAAA", neg =
  "#EE3355"), effect.sizes = NULL, effect.bases = NULL, horizontal = TRUE,
  show.ci = FALSE, ci.prob = c(0.05, 0.95), num.ticks = NULL,
  add.numbers = TRUE, numbers.align = c("center", "left", "right",
  "left_of_bar_end", "right_of_bar_end")[1], numbers.vjust = ifelse(show.ci,
  0, 0.5), left.margin = 0, right.margin = 0, signif.digits = 2,
  round.digits = 8, alpha = 0.8, ...)

Arguments

reg	the results from a regression, e.g. from a call to lm or glm
dat	default = the data frame the regression was estimated from
vars	the explanatory variables that shall be shown in the plot
numeric.effect	a code describing the lowest and highest values of numeric explanatory variables used to calculate the effect, e.g. "05-95" means taking the effect of moving from the 5 percent to the 95 percent quantile.
dummy01	shall numeric varibles that have only 0 and 1 as values be treated as a dummy variables?
sort	if TRUE (default) sort the effects by size
scale.depvar	a scaling for the dependent variable
depvar	name of the dependent variable
xlab,	ylab labels
colors	colors for positive values (pos) and negative values (neg)
horizontal	shall bars be shown horizontally?
show.ci	shall confidence intervals be shown?
ci.prob	left and right probability level for confidence intervals
num.ticks	the number of tick marks on the effect size axis
add.numbers	shall the effect sizes be added as numbers in the plot?
numbers.align	How shall the effect size numbers be aligned: "left","center", "right" align all numbers at the same horizontal posistion for all variables. "left_of_bar_end" and "right_of_bar_end" align them at the end of each bar.
numbers.vjust	Vertical adjustment of the numbers
left.margin	extra margin left of bars as fraction of maximum bar width
right.margin	extra margin right of bars as fraction of maximum bar width
signif.digits	number of significant digits for effect sizes
round.digits	number of digits effect sizes shall be rounded to
...	further arguments passed to qplot. E.g. you can set "main" to specify a title of the plot.

skranz/regtools documentation built on May 30, 2019, 3:02 a.m.