uc_salaries.r

library(XML)
library(ggplot2)

df <- readHTMLTable("http://projects.dailycal.org/paychecker")[[1]]
colnames(df)[4] <- "Salary"
df$Salary <- as.numeric(gsub('[$,]', '', df$Salary))

p <- ggplot(df, aes(x=Department, y=Salary)) + coord_flip()
p + geom_boxplot(aes(color=Rank,
                     x=reorder(Department, Salary, FUN=max))) +
    scale_y_continuous(labels = scales::dollar) +
    labs(title="Salaries by Department",
         subtitle="University of California System",
         y="Annual Salary (2015)",
         x="Department",
         caption="Source: http://projects.dailycal.org/paychecker/\n by @avyfain, inspired by @johnjhorton") +
    theme(plot.caption = element_text(size=7.5))

Came across your tweet about this from 2 years ago. Thought I'd see how to duplicate it in python. Got pretty close, plotnine (python version of ggplot) doesn't do seem to render the caption correctly. This code was tested in python 3.8

# Read the data
import requests
import pandas as pd
html = requests.get("http://projects.dailycal.org/paychecker")
df = pd.read_html(html.text)[0]

# Parse Salary
def money_to_float(money: str) -> float:
    return float(money.replace("$","").replace(",",""))
df['Salary'] = df['Salary (2015)'].apply(money_to_float)

# Plot it. Plotnine doesn't have a reorder function, but using department_list as workaround
from plotnine import *
department_list = list(df.groupby('Department').max().sort_values('Salary').index)
( ggplot(df) 
  + coord_flip() 
  + geom_boxplot(aes(x='Department', y='Salary', color='Rank'))
  + scale_y_continuous()
  + scale_x_discrete(limits=department_list)
  + labs(title="Salaries by Department",
         subtitle="University of California System",
         y="Annual Salary (2015)",
         x="Department",
         caption="Source: http://projects.dailycal.org/paychecker/\n by @avyfain, inspired by @johnjhorton")
  + theme(figure_size=(7, 12), axis_text_x=element_text(size=7.5))
)

That's great @hohonuuli! I did not know about plotnine but thanks for testing it out in Python, that's actually my usual language, but but I couldn't figure out how to do the grouped boxplot properly from pandas, so I went back to R 🙃

FYI, in newer versions of Pandas you can read the url directly into a df: https://stackoverflow.com/questions/32400867/pandas-read-csv-from-url/41880513#41880513

@avyfain Hey, glad it's actually useful for you. BTW, I tried reading the URL directly with df = pd.read_html("http://projects.dailycal.org/paychecker") but it would return a 403 error. I didn't dig into the why though; I just fell back to using requests instead.