sjp.int linetype only in legend, not in graph

[LisaRosenberger]

I am trying to change the linetype of an interaction plot (with a factor and a continuous predictor) into a solid line and a dashed line by changing the linetype in sjp.setTheme():
sjp.setTheme(panel.bordercol = "white",axis.linecolor = "black",panel.gridcol="white",geom.linetype = c(1,3),base=theme_bw())
sjp.int(m10b,type="eff",int.term = "multiplication_factor:pleasantness",show.ci = TRUE)

However the linetype only changes in my legend, and not in the plot itself. Any idea why that is the case?

[strengejacke]

If you use geom.colors = "bw", the line type is applied to the plot, however, it uses black colors (see pkg vignette).

Do you want coloured geoms and different line types?

[LisaRosenberger]

Ah, I didn´t see the vignette, sorry. So using the geom.colors = "bw" does the trick:

However, what I was actually trying to make is a grey scale plot so that the CI´s are more distinguishable, like this:

and there the linetypes didn´t change either. But now I know that it only works with black-and-white plots. And (like in colored plots) there is no actual need for different linetypes with greyscale plots.
Thanks for your quick reply!

[strengejacke]

There is a workaround, by using the returned plot object. sjp.int() returns a list of plots. You can then simply add another fill-scale:

library(ggplot2)
library(RColorBrewer)
library(sjPlot)

fit <- lm(weight ~ Diet * Time, data = ChickWeight)
p <- sjp.int(fit, type = "eff", geom.colors = "bw", show.ci = T)
# 4 colours for 4 CI-bands
p$plot.list[[1]] + scale_fill_manual(values = RColorBrewer::brewer.pal(4, "Greys"))

[LisaRosenberger]

This works perfectly, thanks!

[cathytao]

How to solve the same problem, if using plot_model? I am using the following code and hope to have line dashed/solided across different levels. I did not find a proper solution!

plot_model(p8, type = "int", transform = "exp", terms = "x", geom.colors = "bw", pred.type = "fe", mdrt.values = "meansd")

[cathytao]

Sorry, wrong code -

plot_model(p8, type = "int", transform = "exp", terms = "x", colors = "bw", pred.type = "fe", mdrt.values = "meansd")

[strengejacke]

Either use ggeffects

library(ggeffects)
data(efc)
fit <- lm(neg_c_7 ~ c12hour * barthtot, data = efc)
ggpredict(fit, terms = c("c12hour", "barthtot")) %>% plot(colors = "bw")

or plot_model() with type = "pred" or type = "eff":

plot_model(fit, type = "pred", terms = c("c12hour", "barthtot"), colors = "bw")
plot_model(fit, type = "eff", terms = c("c12hour", "barthtot"), colors = "bw")

For type = "int", it looks like the color-argument is currently not passed down to plot().

[strengejacke]

Should work now with the latest commit to GitHub.