Ordering in geom_dotplot with 2 variables - r

Question

I am using dotplot to do an analysis with two y variables per x variable. I'd like to arrange the chart so that it descends by one of the y variables. I used the reorder() function in the aes() and it reorders it slightly, but not entirely. Chart 1 is what it looks like before, and chart 2 is what it looks like after I use reorder().
Chart 1:
Chart 2:
Here's the code:
answers %>%
ggplot(aes(x = reorder(locale, -percent) , y = percent, fill = box)) +
geom_dotplot(binaxis='y',
stackdir='center',
dotsize = 1,
binwidth = 0.01) +
geom_errorbar(aes(ymin = ci_lo, ymax = ci_hi), width = .5, position = position_dodge(0))
And this is what the "answers" df looks like. The two variables being plotted per locale are in the "box" column - there's a top_box and bottom_box row for each locale:

Answer 1

As pointed out in the comments, you do not provide and data, but I think I have an idea on where you're going wrong.
Here is some example data. I'm going to use a modified mtcars for the example where we will look at the min and max weight of the cars by make.
library(tidyverse)
df <- mtcars %>% rownames_to_column() %>%
select(car = rowname, wt) %>%
mutate(car = gsub("\\s.*?$", "", car)) %>%
group_by(car) %>%
mutate(n = n()) %>%
filter(n > 1) %>%
arrange(car,wt) %>%
filter(row_number() == max(row_number()) | row_number() == min(row_number())) %>%
select(-n) %>%
ungroup() %>%
mutate(stat = rep(c("min", "max"), nrow(.)/2)) %>%
spread(stat, wt)
print(df)
# car max min
# Fiat 2.2 1.94
# Hornet 3.44 3.22
# Mazda 2.88 2.62
# Merc 4.07 3.15
# Toyota 2.46 1.84
Here is what the plot for that would look like:
df %>%
ggplot(aes(x = car))+
geom_point(aes(y = max), color = "red")+
geom_point(aes(y = min), color = "blue")
Now lets talk about what you're trying to do. You say that you would like to order by descending on one of your variables.
df %>%
arrange(-max)%>%
mutate(car = factor(car, levels = car))%>%
ggplot(aes(x = car))+
geom_point(aes(y = max), color = "red")+
geom_point(aes(y = min), color = "blue")
or
df %>%
arrange(-min)%>%
mutate(car = factor(car, levels = car))%>%
ggplot(aes(x = car))+geom_point(aes(y = max), color = "red")+
geom_point(aes(y = min), color = "blue")
I think the key here is that you want to arrange the data and then set the factor levels to get the desired output. If your data is not a factor, then ggplot will use alphabetical order. You may need to spread your data in order to use the exact method outlined above.
Update
You could do this without spreading your data, by arranging with two variables.
Here we will modify the data above to long format
df2 <- df %>% gather(measure, value, -car)
Which plots like this
df2 %>%
ggplot(aes(x = car, y = value, color = measure))+
geom_point()
and then we can arrange without spreading
df2 %>%
arrange(-value, measure) %>%
mutate(car = factor(car, levels = unique(car)))%>%
ggplot(aes(x = car, y = value, color = measure))+
geom_point()
or for descending by min
df2 %>%
arrange(desc(measure), -value) %>%
mutate(car = factor(car, levels = unique(car)))%>%
ggplot(aes(x = car, y = value, color = measure))+
geom_point()