Can dwplot manually set colors to allow for color variation within a coefficient group across models

Question

I am using dwplot to plot multiple coefficients across multiple models. p <- dwplot(list(mod1976,mod1980,mod1984))

The plot outputs coefficients colored by model. I want to color the coefficients so that those that are positive and significant are green, negative and significant are red, and all others are grey.

Is this possible?

Answer 1

This was a little harder than I thought; I couldn't figure out how to override dwplot's colour mapping, but with slightly more effort you can draw your own plot matching the one from dwplot, with more control.

set up example

library(broom)
library(dplyr)
library(dotwhisker)
respvars <- c("mpg", "hp", "wt")
names(respvars) <- respvars ## for .id in map_dfr
predvars <- setdiff(names(mtcars), respvars),
ms <- as.data.frame(scale(mtcars))
mm_list <- purrr::map(respvars,
                      ~ lm(reformulate(predvars, response = .x), data = ms))

construct data frame with estimate/CI/significance info

mm_pars <- mm_list |>
    purrr::map_dfr(tidy, .id = "response", conf.int = TRUE) |>
    dplyr::filter(term != "(Intercept)") |>
    mutate(cat = ifelse(p.value > 0.05, "ns",
                        ifelse(estimate > 0, "pos", "neg")))

plot

Unsuccessful attempt to override the colour mapping:

dwplot(mm_list) + aes(shape = model, colour = cat) +
    scale_colour_manual(values = c("grey", "blue", "red"))

A plot that works (these aren't exactly the colours you asked for, but that should be easy to change ...)

ggplot(mm_pars, aes(x = estimate, y = term, colour = cat, 
    ## I'm not sure the group mapping is necessary ...
    shape = response, group = response)) +
    geom_pointrange(position = position_dodge(width = 0.5),
                    aes(xmin = conf.low, xmax = conf.high)) +
    ## add a reference line
    geom_vline(xintercept = 0, linetype = 2) +
    scale_colour_manual(values = c("grey", "blue", "red"), breaks = c("ns", "pos", "neg"))