Getting Started
Let's get a brief overview of this course.
We'll cover the following
There are several different general approaches within statistics (frequentist, Bayesian, information theory, etc.) and there are many subspecies within these schools of thought. Most of the methods included in this book are usually described as belonging to ‘classical frequentist statistics’. Statistics covers a wide range of general approaches, such as frequentist, Bayesian, information theory, and more. There are many subcategories within these schools of thought. Most of the methods discussed in this course are typically classified as part of the classical frequentist statistical approach. However, this approach, along with its widely used probability values (p-value), have come under increasing criticism.
For example, a researcher may sometimes place too much focus on p-values and not enough focus on effect sizes. This can be an issue because effect sizes such as estimates and intervals are directly related to what we measure when we conduct research. For this reason, we try to take an estimation-based approach that focuses on estimates and confidence intervals when possible.
In addition to the importance of the usage of estimates and intervals, we’ll also emphasize the use of graphs to interpret data and present results. The use of a-priori contrasts (comparisons that are planned in advance) is encouraged. It’s also important to avoid the overuse of multiple testing and instead favor a more focused and planned approach.
Finally, at the end of each chapter, we try to summarize both the statistical approach and what it has enabled us to learn about the science of each example. It is easy to get lost in statistics, but for non-statisticians, the analysis should not become an end in its own right, only a method to help advance our science.
What is the ‘new statistics’ of the title? The term is not clearly defined, but it appears to be used to cover a combination of new techniques—particularly meta-analysis—with a back-to-basics focus on estimation-based analysis using confidence intervals
This course employs a hands-on approach to learning. The code required to perform the basic analysis in R presented in the lesson. This course also skips an introduction to R, and immediately begins with some example analyses. If you’re unfamiliar with the R programming language, we recommend you take a look at Learn R from Scratch before starting this course.