I have a ggplot bar chart with 700 bars, I want to a a Pareto line, it works but the y scale for the bars is too small so they don't show up in chart

Question

Thanks to a lot of people I have my charts working being new to R.

I have three charts

plot frequency ordered

plot Pareto overlay

if You look closely you can see the scaled ordered frequency chart is there at the bottom.

 ```{r}
df <- filter(df_clean_distances, end_station_name != "NA" )
d <-df %>% select( end_station_name) %>%
group_by(end_station_name) %>%
summarize( freq = n())
head(d$freq )
dput(head(d))
d2 <- d[ order(-d$freq),]
d2

plot random

```{r}
ggplot(d2, aes( x=end_station_name, y= freq)) + 
geom_bar( stat = "identity") + 
theme( axis.text.x = element_blank()) +
  ylim( c(0,40000))
```

plot freq ordered

 ```{r}
 ggplot(d2, aes( x=reorder(end_station_name,-freq), y= freq)) +    
    geom_bar( stat = "identity") +   
    theme(axis.text.x = element_blank()) +   
    ylim( c(0,40000))+
    labs( title = "end station by freq", x = "Station Name")

Plot with Pareto overlay

```{r}

ggplot(d2, aes( x=reorder(end_station_name,-freq), y= freq)) +    
geom_bar( stat = "identity") +   theme(axis.text.x = element_blank()) +  
ggQC::stat_pareto( point.color = "red", point.size = 0.5) +
labs( title = "end station by freq", x = "Station Name") 
```

dput(head) output

```{r}
> dput(head(d, n=20))
  structure(list(end_station_name = c("2112 W Peterson Ave", "63rd St 
  Beach", 
  "900 W Harrison St", "Aberdeen St & Jackson Blvd", "Aberdeen St & 
   Monroe St", 
  "Aberdeen St & Randolph St", "Ada St & 113th St", "Ada St & 
   Washington Blvd", 
  "Adler Planetarium", "Albany Ave & 26th St", "Albany Ave & 
   Bloomingdale Ave", 
  "Albany Ave & Montrose Ave", "Archer (Damen) Ave & 37th St", 
  "Artesian Ave & Hubbard St", "Ashland Ave & 13th St", "Ashland Ave & 
  50th St", 
  "Ashland Ave & 63rd St", "Ashland Ave & 66th St", "Ashland Ave & 
   69th St", 
  "Ashland Ave & 73rd St"), freq = c(1032L, 2524L, 3836L, 8383L, 
  6587L, 6136L, 18L, 6281L, 12050L, 397L, 2833L, 1875L, 710L, 1879L, 
  2659L, 151L, 112L, 102L, 78L, 8L)), row.names = c(NA, -20L), class = 
  c("tbl_df", "tbl", "data.frame"))
```    

As you can see the pareto plot is working for the right hand scale but the left hand is out of wack by a lot. While there are 3 million rows the scaling on the y axis has reduced the freq to a very small curve along the bottom, it is there on the left just hard to see.

How do I fix the left y axis to limit to about 40,000 so that the frequency curve shows up correctly?

Answer 1

Here is a solution but not with package ggQC, with sec_axis.
The trick is to pre-compute max(freq) and then use it as a scale factor in order to align the two axis. This data preparation code is inspired in this rstudio-pubs blog post.

library(ggplot2)
library(dplyr)

M <- max(d$freq)

d %>%
  arrange(desc(freq)) %>%
  mutate(cum_freq = cumsum(freq/sum(freq))) %>%
  ggplot(aes(x = reorder(end_station_name, -freq), y = freq)) +    
  geom_bar(stat = "identity") +   
  geom_line(mapping = aes(y = cum_freq*M, group = 1)) +
  geom_point(
    mapping = aes(y = cum_freq*M),
    color = "red", 
    size = 0.5
  ) +
  scale_y_continuous(
    sec.axis = sec_axis(~ ./M, 
                        labels = scales::percent,
                        name = "Cummulative percentage")) +
  labs( title = "end station by freq", x = "Station Name") +
  theme(axis.text.x = element_blank())