bubble chart for groups

[Jet Packer]

Hi,

I am quite new to stata and want to create a bubble chart for visualizing frequencies within a group or value characteristic. I have searched the forum a bit and learned that bubble charts can get messy very quickly, but I still think they might be useful for my case.

Here is an example of what I imagine it could look like:

My Inputs are slightly different:

• y is a continous variable
• x is categorical variable (likert scale, 0-10)
• bubbles should indicate the share of participants for the values of x per level of y

I experimented a bit with a twoway scatter with markers, but I have trouble to connect the size of the bubbles to the the frequency of the x values. If I use x as the frequency weight the bubbles just grow with x but not with the frequency of x per level of y, which I expected intuitively and would be my desired result. What am I missing here?

Best,
Jet

[Nick Cox]

You're asking us to comment on why code you don't show us does not work as you expect on data you aren't showing us either. But frequency weights here just mean that the variable you specify is treated literally as if a count. Frequency weights don't mean that twoway counts frequencies for you.

The graph example you cite seems to me a really good example of why bubble charts are a bad idea. Bubbles are inherently ambiguous. How were bubbles encoded? Is the area being used or the diameter? How are bubbles decoded? Last I looked at the literature -- which includes much discussion in cartography -- people vary in how they read off bubble size but a rough average is that people "see" area to the power 0.7, which is wrong either way.

There are four groups, and a measured variable with bin width 5 Euros. That sounds like a case for four histograms or bar charts to me. Your example sounds similar. The raw data might be interesting too!