Confidence intervals over p-values

Every data science influencer will, at one point, make a post about p-values, discussing how people don’t understand them and then attempting to explain them in a confusing way. Instead of obsessing over p-values, I think people should do more confidence intervals. Confidence intervals elegantly incorporate the concept of p-values when you compute it for a test. And, when you plot or write down a confidence interval as opposed to stating the typical p < 0.05, both you and your stakeholders will gain a much better understanding of what happened and what can happen. That too very quickly.

Also, confidence intervals are useful beyond hypothesis testing. Most things you will compute is an estimate that carry error margins. Everyone is more informed on the risks when you present those margins through confidence intervals.