plot_model type="pred" with own title and solid not dashed error bars does not work

[bridgeovertroubledhuman]

I want a graph using plot_model with type "pred" with my own individual title and solid (not dashed) lines on the error bar. However, both don't work - any help highly appreciated.

My current MWE:

if (!require("pacman")) install.packages("pacman")
pacman::p_load( 'ggplot2', 'sjPlot')
update.packages(ask = FALSE)


data(efc)


efc$sexdummy <- NA
efc$sexdummy <- ifelse(efc$c161sex == 1, "Sex_1", "Sex_2")
efc$sexdummy<- as.factor(efc$sexdummy)
table(efc$sexdummy)



efc$carerrelation <- NA
efc$carerrelation <- ifelse(efc$e15relat == 1, "Child",
                            ifelse(efc$e15relat == 2, "Partner",
                                   "Other"))
efc$carerrelation<- as.factor(efc$carerrelation)
table(efc$carerrelation)

#My model
m2 <- glm(sexdummy ~ carerrelation  , data = efc, family = "binomial")


# The plot with title never works, gets me the error 
 plot_model(m2, type="pred",  title = "My title",
                       show.values = T, value.offset = .3, colors = "bw") 

# Only works without a title 
 plot_model(m2, type="pred", # title = "My title",
            show.values = T, value.offset = .3, colors = "bw") 

Thank you for fixing the issue.

For a picture see: https://stackoverflow.com/q/79396777/6473874?sem=2

Sadly, cannot repex it, as I always get this error:

reprex:::reprex_addin()
ℹ Rendering reprex...
Error in reprex_render() :
This reprex appears to crash R. Call reprex() again with std_out_err = TRUE to get more info.
ℹ Rendering reprex...
Error in reprex_render():
! This reprex appears to crash R. Call reprex() again with std_out_err = TRUE to get more info.
Run rlang::last_trace() to see where the error occurred.

[strengejacke]

Thanks for reporting. For now, as a workaround, you could add a title via ggplot:

library(sjPlot)
data(efc)

efc$sexdummy <- as.factor(ifelse(efc$c161sex == 1, "Sex_1", "Sex_2"))
efc$carerrelation <- ifelse(efc$e15relat == 1, "Child",
                            ifelse(efc$e15relat == 2, "Partner",
                                   "Other"))
efc$carerrelation<- as.factor(efc$carerrelation)

#My model
m2 <- glm(sexdummy ~ carerrelation  , data = efc, family = "binomial")

plot_model(
  m2,
  type = "pred",
  show.values = TRUE,
  value.offset = 0.3,
  colors = "bw"
) + ggplot2::ggtitle("My title")

[bridgeovertroubledhuman]

Thank you! While that works for the title, any chance in getting the error bars from dashed to solid? I tried that with ggplot but so far without any luck.

[strengejacke]

I suggest using the ggeffects package, which is internally used in sjPlot. user interface is very similar, but gives you more flexibility regarding plots:

library(ggeffects)
data(efc)

efc$sexdummy <- as.factor(ifelse(efc$c161sex == 1, "Sex_1", "Sex_2"))
efc$carerrelation <- ifelse(efc$e15relat == 1, "Child",
                            ifelse(efc$e15relat == 2, "Partner",
                                   "Other"))
efc$carerrelation<- as.factor(efc$carerrelation)

#My model
m2 <- glm(sexdummy ~ carerrelation  , data = efc, family = "binomial")


?plot.ggeffects
predict_response(m2, "carerrelation") |> plot()
predict_response(m2, "carerrelation") |> plot() + ggplot2::ggtitle("test")

See
https://strengejacke.github.io/ggeffects/articles/introduction_plotmethod.html
https://strengejacke.github.io/ggeffects/articles/introduction_plotcustomize.html

[bridgeovertroubledhuman]

Oh, I See, I thought it was "just" a small bug in plot_model I had discovered. Thanks for the workaround!