CODEWITHSUNDEEP
rggplot2sjplot
Tomer Zipori
Tomer Zipori

Reputation: 31

Changing a geom_ribbon fill color

I have a mixed effects linear model and I want to plot it using the sjPlot package. I have added a geom-ribbon layer for the CI of each reg line, but I just can't figure how to change it's fill color. Idealy I'd want each ribbon to be the same color as it's reg line.

Here is my code for the model:

model <- lme(no_stickers_gave_stranger ~ s_negativeAff*parent_model,
             data = table4, random = ~1|ifam)

And here is the code for the plot:

plot_model(model = model, line.size = 0.71, type = "int",
           colors = sjplot_pal(palette = "metro"),
           axis.title = c("Negative Emotionality",
                          "Number of sticker shared with a stranger"),
           title = "Age 6 - LAB-TAB",
           legend.title = "Parent Model") +
  scale_fill_sjplot(palette = "metro", discrete = T) +
  geom_ribbon(aes(ymin = conf.low, ymax = conf.high), colour = NA, alpha = 0.25)

The best I could do is this:

enter image description here

Upvotes: 3

Views: 2675

Answers (1)

chemdork123
chemdork123

Reputation: 13793

For whatever reason, setting colors= within plot_model() was not working properly for me either. The way around here is to set your color values the same way you're setting the fill values using either scale_color_sjplot() or scale_color_manual(values=sjplot_pal()).

Curiously, the two methods both work (to match color and fill aesthetics), but give you different actual colors. I have no answer for why, but this is how to match color and fill. I'm using the example dataset from vignettes on the sjplot site.

library(sjPlot)
library(ggplot2)
data(efc)

fit <- lm(barthtot ~ c12hour + neg_c_7 + c161sex + c172code, data = efc)
cols <- sjplot_pal(palette = "metro")

p <- plot_model(fit, type = "pred", terms = c("c161sex", "c12hour"))
p

enter image description here

Just adding the scale_fill_sjplot() function changes fill alone:

p + scale_fill_sjplot()

enter image description here

Adding scale_color_sjplot() fixes that:

p + scale_fill_sjplot() + scale_color_sjplot()

enter image description here

Strangely, if we use the scale_*_manual() functions and supply the sjplot_pal() as the reference for values=, we get different colors, but at least they do match:

p + scale_fill_manual(values=cols) + scale_color_manual(values=cols)

enter image description here

Without knowing more bout the sjPlot package and plot_model(), I would guess the difference in color has to do with the package assigning levels and colors a bit differently than ggplot2 does by default (which is either alphabetical or based on the levels() if you have a factor).

Upvotes: 1

Related Questions