At the recent 2019 Canadian Psychological Association convention, I gave a workshop on Cloud-Based Statistics Apps that could potentially be useful for teaching. I’m posting the list of applications here in case anyone finds them useful for teaching. In addition, also posting my slides from the talk, where I describe my approach to using these apps when teaching undergrad statistics.
Central Limit Theorem (Means)
This application runs simulations to demonstrate the central limit theorem from different population distributions by sampling up to 1000 times from a population. It shows the population distribution, the first 10 sample distributions, and the sampling distribution. This apps helps enormously with teaching students the central limit theorem. It can sample from normal, uniform, right-skewed, and left-skewed population distributions.
Central Limit Theorem (Proportions)
This app is like the one above, except that it shows the sampling distribution for proportions.
This app simulates new data for each run and shows students how to review diagnostic plots from linear regression models. It allows students to simulate situations where the normality assumption is likely to be met as well as situations where the assumptions are violated (quadratic relationships, non-constant variance). Helpful for getting students to interpret plots of residuals.
Generic Distribution Calculator
Calculates area under the curve for various distributions, such as z, binomial, t, F and χ2. Useful when teaching students about standardizing distributions (e.g., converting to z-scores) and for conceptual understanding of p-values.
Note that Bruce Dudek does not post his code for these apps online. However, he may be willing to share code with you if you send an email (I use some of these apps in my own classes!).
Distributions (z, t, F & chi-squared)
Bruce developed a set of distribution apps that improve upon the generic ShinyEd one above in some ways. These apps are primarily useful for looking up p-values, and can replace looking up p-values in conventional critical value tables from old-school statistics textbooks. If the emphasis is entirely on looking up p-values, these apps work better.
Descriptive Statistics and Visualization
This is a cool app that has a few datasets built in, and allows students to calculate descriptive statistics (e.g., means, SDs, range) as well as a few basic data visualizations (e.g., histograms, box plots). Can be useful as a way to introduce students to exploring data with graphs.
This app simulates confidence intervals (up to 100 at a time). It is useful to help students understand the conceptual issues surrounding confidence intervals, and the impact of sample size and confidence limit (e.g., 90% vs 95%) changes the width of the intervals.
RShiny Apps from Other Sources
This app has a LOT going on. It can allow the user to simulate data and run a one-way ANOVA, complete with beautiful data visualizations. It can also allow the user to read in data to run a one-way ANOVA on a pre-specified dataset. Finally, it also allows the user to run individual t-tests as a rudimentary form of post-hoc test. This app is very useful to help students understand and interpret one-way ANOVA.
Other Online Apps that Don’t Use RShiny
The applications below don’t actually use RShiny but are similar in the sense that they can be run in a web browser without having to download software. I’ve found them to be useful in teaching.
Guess the Correlation
A game that shows students a scatterplot and asks them to guess the magnitude. Very fun.
This allows you to add dots to a scatterplot with a point-and-click format, calculates the correlation coefficient and allows you to save the resulting dataset. Insanely useful for teaching.
Rock ‘n Poll
Explains the concept of sampling variability using political polling as an example. Very beautiful viz.
Rossman/Chance Java Applet Collection
Kristoffer Magnusson’s d3.js Visualizations
Kristoffer has made some truly excellent data visualizations worth checking out. The ones above cover null hypothesis significance testing, confidence intervals, effect size, t-distributions, and equivalence testing. I honestly can’t hype these enough, they are all top-notch.
Online cloud-based statistics software
Sometimes, you want your students to analyze some data in a similar fashion to how they would if they had conventional software installed (e.g., SPSS, SAS). All of these do all of the calculations online, and remove the need for students to install software – which, I have discovered, is a very error-prone process for many students. Useful if you want students to do statistical calculations in a class where learning a particular software is not the goal.