Abstract

Physics education research studies are prone to omitted variable bias because we never know the ``true'' model and can never account for all possible variables.

Highlights

  • Quantitative research in physics education often involves comparing outcome measures across other variables including modes of instruction and student characteristics

  • We instead built models based on hypothetical causal models that are not based in any theory or prior work, which allowed us to demonstrate the effect of omitted variable bias

  • We can interpret this coefficient through the following: Given two students, one who intended to major in physics and another who intended to major in engineering, with identical E-CLASS pretest scores and provided the same type of lab instruction, we would expect, on average, the student who intended to major in physics to score 1.0 points higher on the post-test than the student who intended to major in engineering

Read more

Summary

INTRODUCTION

Quantitative research in physics education often involves comparing outcome measures across other variables including modes of instruction and student characteristics. We may implement a new pedagogy in one semester and evaluate student performance on its own or evaluate survey data to understand students’ experiences with the new pedagogy Both of these types of studies can be used to make causal claims, but require stronger assumptions about how data were. For example, found that students’ scientific reasoning ability [8,9,10,11,12], their prior preparation [13,14,15], their self-efficacy [14], their major [16], or biases inherent in the assessment instruments [13] are confounding variables in analyses of students’ performance on standardized assessments In each of these cases, performance differences between conditions or subgroups of students previously observed were instead at least partially explained by considering an additional, previously omitted variable. Throughout, we consider the following hypothetical research question: What is the causal effect of a students’ intended major on their beliefs about experimental physics?

Methodological considerations
CONDITIONS FOR OMITTED VARIABLE BIAS
ILLUSTRATIVE EXAMPLES OF OMITTED VARIABLE BIAS WITH REAL DATA
Data sources
Methods
Model 1
Model 2
Model 3
Summary
DISCUSSION
CONCLUSIONS
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call