Racer is a game where the player can race a variety of racecars on multiple tracks. Each car body, engine and set of tires impacts the speed and acceleration of the cars. Players can choose any options they like, but they will develop better strategies (i.e. improve their chances of winning) if they thoroughly understand the data and techniques for analyzing it.


Racer Lab Part A Making Decisions with Data: Student Handout (Instructor Notes)

Part A of this lab is designed to encourage students to carefully consider how to make with decisions with data. In steps 1 and 2, students do not need any familiarity with hypothesis testing.

The goal of this lab is to emphasize the importance of carefully evaluating the data before any conclusions are drawn.

Sample Tutorial Data: Example Data

Cleaned Sample Tutorial Data: Cleaned Example Data


Racer Lab Part B Making Decisions with Data: Student Handout (Instructor Notes)

Part B assumes students have been exposed to paired and two-sample t-tests. This activity uses the Racer game to demonstrate the challenges in drawing conclusions in hypothesis tests. Students recognize that multiple sources of bias and confounding variables can cause the p-value to be unreliable. The emphasis is on using data to draw conclusions, even when the standard assumptions for hypothesis tests may not hold.

Sample Oval Track Data: Example Data


Racer Lab Designing Experiments: Student Handout

This lab is independent of the other two labs. It allow students to formulate their own hypotheses, collect data, conduct an analysis, and draw conclusions. These labs are flexible so that students can select from a variety of explanatory and response variables based upon their own unique research question.


The following link allows you to play the Racer Game. (Click on the Racer tab). These games may not run on the most recent versions of Chrome or Firefox. We are currently working on fixing the issue.


You may be asked to install Unity Web Player, this may take a few minutes.

Many browsers will require you to allow popups before they will run these stat2labs games.


Data visualizations for all Racer game is available on the Racer Visualizations App.


All data from the game is available at Racer Data.


Learning Goals

  • Discuss practical challenges in collecting meaningful data (sample size, bias, when it is possible to get a true simple random sample, time constraints and other resource limitations).
  • Using an interactive data visualization app to identify challenges with real (i.e. messy) data.
    • Recognize the prevalence of errors in data (essentially all data is messy)
    • Recognize the importance of reproducible research (documentation of data collection, data cleaning and analysis are all essential components of data-based decision-making)
    • Provide clear examples of sampling variability (while textbook data is fixed, all samples from a population will vary)
  • Provide a structured approach to gathering experimental data. This involves understanding variability and isolating key factors of interest in experimental design, such as:
    • selecting appropriate response variables, factors and levels,
    • addressing unwanted variability (restricting the population of interest, blocking, and adding additional variables or interaction terms into a model)
  • Explore and analyze data using appropriate graphs and numerical tools.
  • Derive appropriate, actionable conclusions from data. This involves training students to articulate the core differences between calculating a p-value and data-based decision-making.
    • How have we addressed violations in the statistical assumptions?
      • Is there evidence we are violating the normality (or other model) assumptions?
      • Is the sample size large enough?
      • Does the sample data properly reflect the population? Is it a true simple random sample of the population?
      • Are there any reasons to believe the data may be biased based upon how the data was collected?
    • When can we make conclusions about associations between variables? When can we use sample data to show evidence of causality?
    • How are we ensuring full reporting and transparency?
    • Are there data points that should be removed? Are there missing values that may affect the validity of the data?
    • Recognizing the difference between small p-values, the size of an effect and the importance of the result.
    • Recognizing that a p-value should never be the singular measure of evidence regarding a model or hypothesis.


Thanks to Grinnell students Gemma Nash, Mariam Nadiradze, Anaan Ramay, Dev Nalwa, Ryuta Kure, Reina Shahi, Priyanka Dangol, and James Msekela for creating, editing and maintaining the on-line game.