I wanted to include a plot_summs output in a paper and was unhappy with how thin the CI lines were. I therefore added custom functions to my code that are based on your plot_summs and plot_coefs. I'm only a novice programmer busy with other things, so I don't really have the skill or time to adapt my changes to all possible types of plots (eg I only worked without distributions) or to properly test them, but I still thought I should let you know what I did, maybe as a starting point if you want to implement this feature at some point.
I simply passed a new argument called line.width to ggplot in the same way as point.size, though I multiply line.width with 0.8 to make sure an input of 1 gets the same thickness as the current default.
Then I changed fatten and size in the following section, and added the guides() line to undo the fattening in the legend
line.width <- line.width * 0.8
# If there are overlapping distributions, the overlapping pointranges
# are confusing and look bad. If plotting distributions with more than one
# model, just plot the points with no ranges.
if (plot.distributions == FALSE || n_models == 1) {
# Plot the pointranges
p <- p + ggstance::geom_pointrangeh(
aes(y = term, x = estimate, xmin = conf.low,
xmax = conf.high, colour = model, shape = model),
position = ggstance::position_dodgev(height = dh),
fill = "white", fatten = point.size / line.width, size = line.width,
show.legend = length(mods) > 1) + # omit legend if just a single model
guides(colour = guide_legend(override.aes = list(size = point.size/4)))
} else {
p <- p + geom_point(
aes(y = term, x = estimate, colour = model, shape = model),
fill = "white", size = point.size, stroke = 1, show.legend = TRUE)
}
I wanted to include a plot_summs output in a paper and was unhappy with how thin the CI lines were. I therefore added custom functions to my code that are based on your plot_summs and plot_coefs. I'm only a novice programmer busy with other things, so I don't really have the skill or time to adapt my changes to all possible types of plots (eg I only worked without distributions) or to properly test them, but I still thought I should let you know what I did, maybe as a starting point if you want to implement this feature at some point.
I simply passed a new argument called line.width to ggplot in the same way as point.size, though I multiply line.width with 0.8 to make sure an input of 1 gets the same thickness as the current default. Then I changed fatten and size in the following section, and added the guides() line to undo the fattening in the legend
Hope it can be of some use :)