One Crank or Two?

• Activity Number: 651
• Type: Contributed
• Date/Time: Thursday, August 7, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Education
• Author(s): Kari Lock Morgan*+
• Companies: Duke University
• Keywords: randomization test ; resampling ; simulation ; introductory statistics

Abstract

Students should learn to distinguish between random sampling from the population and random assignment to treatment groups in an introductory class. When teaching inference with randomization tests, should we teach different types of randomization, depending on the source of "randomness" in data collection? Do we want our students to analyze data from observational studies differently than data from randomized experiments? This has come to be known as the question of "one crank or two," a phrase used by John Holcomb at ICOTS8 in Slovenia. The "one-crankers" like the simplicity of being able to use the same procedure in all circumstances and believe that introducing two cranks creates unnecessary confusion for students. The "two-crankers" may teach one approach (reallocating) for data from randomized experiments and another approach (resampling) for data from observational studies. In this talk we present both sides of the debate, and give results from simulations comparing analysis via one crank or two.