In the last post, we saw a basic overview of Bayes rule, but we didn’t experience it at its full power. Rather than limiting ourselves to single value priors, we can use a full distribution as a prior. In this post, we’ll see a Bayesian example with an expanded prior and compare it to the more traditional frequentist methodology.

Up until now, most of the statistical work we have looked at would fall under the category of frequentist statistics: we’ve made inferential conclusions about populations based solely on data.

But we don’t necessarily need to rely on just a sample of data to make a conclusion; we can include our prior knowledge and beliefs on the subject using what is called Bayesian statistics. The next couple of posts will focus on Bayesian methodology, allowing us to see how we can include our own prior beliefs in models we create.

Statistical models are a lot like us: in order to become stronger, they have to first bulk up. While my daily diet of handfuls of birthday cake flavored protein powder hasn’t yet paid off in the strength department, it’s no mystery what a model needs to get stronger: data.

Power analyses can give us answers to questions of how much data we need to identify a certain size effect in an inferential model. Come with me, as we judge the proverbial statistical model Mr. Universe contest.

Up till now, we’ve gone over a lot of different topics, some complicated, and some not so much. In this tutorial, I want to go over a collection of more basic ideas that may not require their own post, but are helpful to look back on and get a better understanding of. Yes, putting on an aire of knowledge is fun and makes us look smart (and in some circles “cool”), but actually knowing what we’re talking about gives us even MORE credibility. Take off the temporary mask of faux-know-it-all-ism and get the permanent facial reconstruction of actual know-it-all-ism in a series I like to call, “Back to Basics”.

Previously in these blog posts, we’ve looked at predicting some variable using a bunch of available statistics. In this post, we’ll look at using time to predict a value. Time series models allow us to make predictions using a more flexible framework for… time series data (crazy, I know)! There are many time series models, each with their own pros and cons, but in this post, I’m going to focus on ARIMA models and their components.