Tutorials to accompany Stats for Data Science

An important question for statistics instructors is how to incorporate computer technology into their course. I say “how” and not “whether” because:

  1. Contemporary statistics is done with software. This includes basic data handling and graphics, as well as carrying out hypothesis tests and statistical modeling.
  2. Introducing ideas like sampling variation is readily done with technology such as resampling.

Still, an instructor has to deal with the realities of her and her students’ situation. Are the students fearful about computing? What computing facilities (hardware and also software) are available to students? How much of the course is the instructor willing to turn over to learning how to carry out computations? How comfortable is the instructor carrying out a computaton in front of students in the face of innumerable software issues, the possibility of making a small mistake that makes your demonstration fall apart? Answers will vary from one instructor to another.

Stats for Data Science incorporates three different “levels” of computing. This lets an instructor select an approach to computing that matches the situation she faces.

  1. Little Apps are web-based, mouse-driven demonstrations that require no software set-up and involve zero coding by students. There are several thoughtful groups that have published freely available apps for statistics. Among these are:

    What’s distinctive about the Little Apps is:

    1. They are always based in a graphical display of real data.
    2. The data sets tend to be large (e.g. n = 10,000) and with many variables (say, 10 to 50)
    3. They have a highly consistent user interface. For instance, every Little App has the same controls to select a data frame and to choose the response and explanatory variables and the sample size.
    4. They are not directly about statistical theory. Instead, they are designed so that students can explore (or be lead in exploring by the instructor) statistical concepts using data, and the ability to draw new samples and (occasionally) to randomize or bootstrap.

    You can access the Little Apps here. On that same site you’ll find a set of about 20 activities based on the Little Apps, as well as instructor notes about central topics in a conventional Stat 101 course.

  2. R Tutorials are also web-based and do not require any software set-up. Every tutorial combines a narrative about the subject with interactive blocks containing R-language statistical calculations. Rather than starting with a blank slate, and having to figure out the details of the computation by trial and error, the interactive blocks almost always have working code that can be run with a button click. Students can edit this code, which can be a matter as simple as changing the name of a variable, a sample size, and so on. The interactive blocks are usually based on mosaic package R commands, which offer a very consistent interface to carrying out a variety of statistical operations including graphics and modeling. The best way to see what the tutorials are about is to try one. If you have no background in R, start with the beginner’s tutorials. You’ll quickly learn the basic syntax of commands and then can jump to the tutorials that lay out the computational techniques relevant to each chapter of Stats for Data Science.

    Beginner’s tutorials

    Statistical tutorials

  3. R Markdown

https://dtkaplan.shinyapps.io/SDS-bayes/

https://dtkaplan.shinyapps.io/SDS-effect-size/

Mouse driven computation that allows you to avoid coding while providing considerable power in summarizing data and exploring statistical concepts