The lead chart of Le Baguette Index is a dot plot with some unusual jitter. Each dot is different bakery in France and the X axis is the price of a baguette.
At first, I didn’t realize there was any x-jitter and planned to try adding some. But after downloading the data, I could see the complication. There are 1640 bakeries sampled, and many of them use the same round-number prices, such as 1.3€. Here’s a standard dot plot with minimal x-jitter using flat “dots” to reign in the height a little.
I call it a “pancake chart“. I’ve also added color for the cities with the most bakeries, to reenforce that the chart is stacks of individual marks and not a bar chart.
The pancake chart is still uncomfortably tall, so I can understand the need to explore more aggressive x-jitter methods. It looks like the original added some random-uniform jitter in the x direction, possibly wider for the more popular prices, which allows for a constrained height. That’s probably necessary for a dot-based view, but first, here’s another pancake chart, but this time allowing maximal smoothing across pancake stacks so the heights follow the overall density better.
Not bad, though the pancake shape makes it hard to see the outlying prices.
I tried the variable-width x-jitter idea, except using gaussian random instead of uniform random to keep the kernel density spirit. In the original, the less common prices like 1.5€ still get a good bit of x-jitter, so it’s not a straight proportional adjustment to the jitter width. Here’s a square root proportional adjustment, which is enough to allow dots instead of pancakes.
After adjusting the prices, I used a hex grid dot plot with minimal smoothing, since the adjustment was already doing a lot of smoothing. Maybe that blurs the discreteness of the prices too much. Here’s a dot plot with simple uniform x-jitter along with dot plot smoothing.
It’s nice that the stacks are separated enough to still suggest the discreteness of the prices.