Teaching Statistical Inference via Simulation using R

Abstract

George Cobb describes the core logic of statistical inference in terms of the three Rs: Randomize, Repeat, Reject. (See repositories.cdlib.org/uclastat/cts/tise/vol1/iss1/art1) Note that all three Rs involve process or action. Teaching this core logic is more effective when students are able to carry out these actions on real data. In this webinar, I'll show how to use computers effectively with introductory-level students to teach them the three Rs of inference. To do this, I will use a another R: the statistical software package. The simulations that will be carried out involve constructing confidence intervals, demonstrating the idea of "coverage," hypothesis testing, and confounding and covariation. Although R is professional-level software, it's very easy to use in an introductory setting, as I have been doing for the last decade. The key is to use flexible and concise operators. I'll provide these to the seminar participants. To follow the seminar successfully, you do NOT need to know anything about R or computer programming. However, you should install R on a computer so that you can follow along. Instructions for doing this, and a short introduction to simple R commands, are available at www.macalester.edu/~kaplan/ISM/draft-intro.pdf (PDF) (see Section 1.4).

A note from the presenter:
Dear Webinar Participants,

Here are the slides for next Tuesday's webinar. I'm sending them out in advance because they contain information on how to install R and the datasets, etc. for the webinar. The slides also contain some background and extension material that there won't be time to go over during the webinar --- these are the slides marked with a dark band at the bottom.
I'm also attaching a "crib sheet." Although there are just a few commands that you will need to learn to carry out the simulations described in the webinar, it's convenient to have these all listed on one sheet.
This is the first time I have given a presentation using web-based software. I have been practicing a bit. One of the things I have realized is how different the webinar format is from the conventional face-to-face situation. In classes and workshops, I have always liked to have students or participants use the computer at the same time as we are talking about the statistical principles and how the computer supports them. Inevitably, people make mistakes, but these become learning experiences since I am there to help get them quickly back on track.
In the webinar format, however, I have no practical way to see what you are typing in your own R sessions and no way to respond quickly to errors. So, I'm concerned that people who are trying to follow along in their own R session will just get distracted. I suggest that the best way to proceed, if you do get distracted by a small error, is to stop and follow the webinar --- I hate to say it --- "passively." Then, after the webinar, we can sort through any problems in a one-on-one format. I find this regrettable, since I think people learn better when they are actively engaged with the material, and because the basic premise of this webinar is that when students actively implement the logic of statistical inference, they come to a faster and better understanding of it.
Regards,
Danny Kaplan

Watch Webinar Recording (FLASH)