Bayesian methods give more intuitive answers, are easier to work with and conceptualize, and give sensible answers for data sizes as small as N=1.
When using Bayesian methods:
*confidence intervals get easier * the hacky concept of p-values is discarded * Uncomfortable cutoffs like "we can use this method if we have ~30 data points, but not if we have less" disappear.
There are many skilled data scientists who think statistics is confusing and doesn't make sense. They're right: the popular school of statistics, still taught in most universities, is a confusing mess. Bayesian methods cut through the confusion, and make advanced statistics accessible to the average data scientist.
For example, if you find p-values confusing... yes. You should. They are just bad. There is a far better alternative, which I'll tell you about.
Bayes' theorem has been covered to death, so I won't be covering it. Nor will I dump vicious equations on you. Instead, this talk has two aims:
1) To free the listener's mind from the hodge-podge of rules half-remembered from school, replacing them with a theory of probability that makes sense and can be reasoned about.
2) To equip you with methods you can use immediately, right when you get home.
and a bonus:
3) To show you how statistics works (because it absolutely does) when N = 5, or N = 2.
