Specify model order in multiplot() function - r

Question

I am using the multiplot() function in R with the coefplot() function to create a plot with coefficients from several models.
multiplot(mod1, mod2, mod3, mod4, mod5, mod6, mod7, mod8, mod9,
intercept = F, zeroType = 1,
zeroColor = 'black',
title = 'The Effect of Masculinity on Sexism Items',
ylab = 'Masculinity Index',
plot.shapes = T,
xlab = 'Sexism',
sort = 'natural',
legend.reverse = F,
names = c('Control', 'Offend', 'Prob', 'Protect', 'Rescue',
'R/W', 'Equal', 'Hire', 'Harass'),
newNames = c(masc_ind = 'Masculinity'))+
scale_color_manual(values = c('green', 'green', 'green', 'blue', 'blue',
'blue', 'red', 'red', 'red'))+
scale_shape_manual(values = c(15,16,17,18,19,0,1,2,5))
Which produces the above plot:
I noticed that the order of the models does not follow the order they are entered into the code (ie: mod1, mod2, mod3). Instead, the models appear to be listed in reverse alphabetical order. I know there are some ways to override this in ggplot() for factor level variables in a single model, but I haven't been able to find much in the way of reordering models in ggplot(). Is there a way to override this so I can list the models in a preferred order?

Answer 1

The difficulty with ggplot wrappers like multiplot is that often what you gain in ease-of-use, you lost in flexibility. However, the data frame that multiplot produces for its plot data is a very nice little summary table. This can be used to plot directly in ggplot however you choose.
Obviously, we don't have your data, but I have created a data set with all the same names and structure as your own (see below), so the following code should work for you:
library(coefplot)
var_names <- c('Control', 'Offend', 'Prob', 'Protect', 'Rescue',
'R/W', 'Equal', 'Hire', 'Harass')
plot_data <- multiplot(mod1, mod2, mod3, mod4, mod5, mod6, mod7, mod8, mod9,
intercept = F, zeroType = 1,
zeroColor = 'black',
title = 'The Effect of Masculinity on Sexism Items',
ylab = 'Masculinity Index',
plot.shapes = T,
xlab = 'Sexism',
sort = 'natural',
legend.reverse = F,
names = var_names,
newNames = c(masc_ind = 'Masculinity'))$data
p <- ggplot(plot_data, aes(Value, factor(Model, rev(var_names)), color = Model)) +
geom_vline(xintercept = 0) +
geom_errorbarh(aes(xmin = LowOuter, xmax = HighOuter), height = 0) +
geom_errorbarh(aes(xmin = LowInner, xmax = HighInner), height = 0, size = 1) +
geom_point(aes(shape = var_names), size = 4) +
labs(x = "Sexism", y = "Masculinity Index") +
theme(legend.position = "bottom")
p + scale_color_manual(values = setNames(c('green', 'green', 'green', 'blue',
'blue', 'blue', 'red', 'red', 'red'),
var_names), name = "Variable") +
scale_shape_manual(values = setNames(c(15, 16, 17, 18, 19, 0, 1, 2, 5),
var_names), name = "Variable")
And of course you can style this however you choose:
p +
theme_minimal(base_size = 16) +
scale_color_manual(values = setNames(rep(c("red3", "orange2",
"steelblue4"), each = 3), var_names),
name = NULL) +
scale_shape_manual(values = setNames(c(15, 16, 17, 18, 19, 0, 1, 2, 5),
var_names), name = NULL) +
labs(x = "Sexism", y = "Masculinity Index") +
theme(legend.position = "bottom")
Data used
set.seed(1)
var_names <- c('Control', 'Offend', 'Prob', 'Protect', 'Rescue',
'RW', 'Equal', 'Hire', 'Harass', 'masc_ind')
df <- setNames(as.data.frame(matrix(rbinom(500, 1, 0.5), 50)), var_names)
list2env(setNames(lapply(head(var_names, -1), function(x) {
df2 <- data.frame(masculinity = df[[x]], masc_ind = df$masc_ind)
do.call("glm",
list(formula = masc_ind ~ masculinity,
data = quote(df2),
family = quote(binomial)))
}), paste0("mod", 1:9)), globalenv())