Bayesian analysis for a Class of Beta Mixed Models

Over the past decade, interest in the generalized linear mixture model (GLMM) has increased. One reason for this popularity is that GLMM combines the Generalized Linear Model (GLM) \ citep {Nelder 1972} with the Gaussian variate effect, enhances the flexibility of the model, makes it complicated data structures such as stratification and iterative measurements It is to adapt. Such as longitudinal direction, often result in additional variability and / or dependency. GLMM can also be regarded as a natural extension of the mixed linear model \ citep {Pinheiro: 2000}, allowing for flexible distribution to response variables.

In the last semester of the university, I did independent research on data mining. This course covers three extended teaching materials: Hastie, Tibshirani, Witten, James, Bayesian data analysis (Kruschke), and time series analysis and application (Shumway, Stoffer). We conducted a lot of exercises in Bayesian analysis, Markov chain Monte Carlo, hierarchical modeling, director, and unsupervised learning. This experience has confirmed me to deepen the interest in the disciplines of data mining and further focus on it. Recently, I have completed the online course of Stamford Lagunita 's statistical study. This covers all the materials of the book "Introduction to Statistical Learning" that I read in an independent study. Since I mention the content twice, I would like to share 10 statistical methods in this book. I think that any data scientist should learn to handle big data sets more effectively.

The first focal region in applying Bayesian reasoning is Bayesian linear modeling. In this article I am trying to introduce the concept of Bayesian linear regression. We will review the frequency linear regression method easily, introduce the Bayesian interpretation, and look at some of the results applied to a simple data set. I left the code of this article, but it is in the GitHub of the Jupyter notebook.

Even after struggling for weeks on Bayesian linear model theory and blog posts, I can not say that I completely understand the concept. Therefore, the way of thinking through practice is the most effective technology. I started using Bayesian linear regression as my selected machine learning model for data science project. This article is the first of two articles that documented the project. I'd like to show you an example of a complete data science pipeline, so the first article will focus on problem definition, exploratory data analysis, and benchmark setup. In Part 2, we will focus on completely achieving Bayesian linear regression and interpreting the results. Therefore, if you already have EDA, proceed. Otherwise, or just want to see a good plot, stay here. I will show you how to start with the problem of data science.