1

I'm doing some plotting of some mixed models I've been running and am pulling my hair out trying to make sjp.lmer() change the x and y axis labels of a fixed effect. If I'm missing something simple, please let me know!

This is my code:

library(sjPlot);library(lme4)

model = lmer(DV ~ IV + (1|groupingVariable), data = data, REML = F)

sjp.lmer(model, 
     type = "fe.slope", 
     vars = c("IV"),
     title = "Estimated effect of IV1 on DV", 
     geom.colors = c("black", "grey49"), 
     show.ci = T, 
     axis.title = c("IV Title", "DV Title"))

Model is a formal model estimated using glmer(). The problem is that regardless of what I write, the x and y labels do not change.

I think the syntax is correct, because this code works:

sjp.lmer(model, 
     type = "re", 
     sort.est = "sort.all", 
     facet.grid = F,
     axis.title = c("IV Title", "DV Title"))

This second being a plot of the random effects for the same model. Is this a bug? Can you for some reason not specify axis labels for fixed effect models? Thank you!

Seems like a bug. I've raised the issue on gitHub at:

https://github.com/sjPlot/devel/issues/212

Cole Robertson
  • 599
  • 1
  • 7
  • 18

1 Answers1

2

As far as I can tell it's just an oversight ("misfeature"/bug). Looking at the code here:

reglinplot <- reglinplot +
  labs(title = title,
       x = sjmisc::get_label(model_data[[p_v]], def.value = p_v),
       y = response)

it seems the labels are hard-coded. Furthermore, the argument list of sjp.reglin (here) doesn't have an axis.title argument ... posting an issue, as you have done, seems like the right way forward.

However, it's not too hard to hack the plot a little bit if you know just a tiny bit about the ggplot2 package.

Set up example:

library(sjPlot); library(lme4
mod <- lmer(Reaction~Days+(Days|Subject),sleepstudy)
p1 <- sjp.lmer(mod, 
     type = "fe.slope", 
     vars = "Days")  ## stripped-down (warning about colour palette)

Hack labels:

library(ggplot2)
p1$plot.list[[1]] + labs(x="hello",y="goodbye")
Ben Bolker
  • 211,554
  • 25
  • 370
  • 453
  • axis-titles are not set for this plot-type, because you can have multiple plots (see `plot.list` return value), which all have different x-axis-titles (the name for the IV, for instance), while the y-axis is always the same title (the DV). That's why I did not implement any option for setting axis titles, and indeed the workaround is to use your suggestion to hack the labels. – Daniel Feb 28 '17 at 18:28
  • Perhaps the issue becomes a documentation request then ... ? – Ben Bolker Feb 28 '17 at 20:04
  • That would be the easiest solution, and probably even easier for the user than dealing with a complicated list of values as `axis.title`-argument. I'll think this over... – Daniel Feb 28 '17 at 21:55