We recently published an analysis of the 403 episodes of PBS’s “The Joy Of Painting,” 381 of which were done by the inimitable Bob Ross. We’ve made the raw data and our clustering code available on GitHub for those who want to play with the data at home.
Some have already taken a stab at it. Here are a few of our favorites.
Jesse Paquette, who created the cancer-research tool EGAN and is developing software for tag-based analysis, sent us this visualization of Ross’s paintings with fences:
And Emre Barut, who has done work in high-dimensional modeling, fit another clustering model, an Isling model, to our set. “Each node is a different object,” he said in an email, “and the edges between nodes represent a level of correlation. Positive correlation is green, and negative is red. The edges’ widths also represent the strength of the correlation.”
Brian Keegan, who has checked our work before, also tried his hand at our data. His analysis of the set is worth a read. Here, he used network-visualization tool Gephi to plot and identify sub-communities from our Ross data.
And Tom Anson wrote in to call us out on our chimney reporting. Annette Kowalski, the co-founder of Bob Ross Inc., had relayed an anecdote about now Ross avoided painting chimneys because they were more evidence of people than he preferred to have in his paintings. We found a painting with a chimney, but it was hard to do. Anson claims to have found another:
Interestingly enough, I literally just watched Season 4, Episode 1 (“Purple Splendor”), wherein he paints a chimney. I just thought you would like to know that a chimney appears at least twice (given your already noted Season 7, Episode 1).
Thanks to everyone who chimed in. We’ve made such data available with the goal of people continuing to research and expand the set. So interact with it, fork it and edit irregularities you find. If we missed your happy spreadsheet, shoot me an email.
And remember, we don’t make #DIV/0!s, just happy little accidents.