Analysis of *small* samples

Hello, I am watching the 3rd session of the Program Evaluation equity training. Around 1:22, Heather is discussing the analysis of likelihood to vote by sexual orientation and presenting the uncertainty of estimates for smaller groups. I love this approach but I’m not sure what type of analysis is used to create the estimate and the confidence interval. I would expect there’s a reference group, and if so, what is it? Or if not, what am I missing? THANK YOU!


Hi there @svalen11 great question! I am so glad you asked it here. :slightly_smiling_face:

In order to calculate an estimate and a confidence interval, you don’t need to specify a reference group. You can simply calculate an estimate for each group on its own.

You do need a reference group if you are going to calculate differences in the likelihood to vote. That’s because, in order to compare two or more groups, we need to set one as a mathematical “baseline.” Setting one group as a baseline doesn’t have the same effect as the social norming of one group as it doesn’t change the actual math, but it should still be done with care as the interpretation and narrative around a “baseline” or “reference” group can reinforce privilege.