By Kristen E. Roland and Jennifer J. Kaplan, University of Georgia
Information
Statistical inference is a difficult concept to teach and learn. While there are a number of multiple-choice assessments that evaluate undergraduate students’ knowledge of p-values, constructed-response questions, in which students write an answer in their own words, have been shown to reveal students' understanding better than multiple-choice questions. In addition, the use of constructed-response questions may better address the second recommendation of GAISE: use assessments to improve and evaluate student learning. The findings on student conceptions of p-values presented in this poster are based on a qualitative analysis of student writing about the probability represented by a p-value. The original student data were collected through an online homework system used in a large-lecture format introduction to statistics course (enrollment of ~1,200 students per semester) at a large research institution. These data are being used to create machine learning models that can categorize new student responses to the same questions. This poster will not only provide initial findings about the distribution of conceptions of p-values that arise when students are asked to provide a narrative description of the probability represented by the p-value, but also a website through which instructors can upload students’ responses and receive real-time formative feedback about student interpretations of p-values for one proportion and one mean hypothesis tests. The feedback can then be used as part of just-in-time-teaching. Both the results presented and the feedback provided by the website can be used by instructors as evidence to better educate their students concerning the appropriate interpretations and limitations of statistical inference. Understanding the ideas students form during an introduction to p-values will help teachers modify instruction later in the course to support student understanding, producing more effective teaching and promoting better student learning.