Evaluation of exam questions using bootstrapping: Practical applications in R and SPSS with a case study

Area of Science:

• Educational Measurement
• Psychometrics
• Statistical Analysis

Background:

• Parametric analysis for exam question quality requires large samples, which are often unavailable in medical education.
• Small cohorts can lead to unreliable psychometric metric estimates (e.g., item difficulty, point-biserial correlations) using traditional methods.
• Flawed assessments may result from unreliable metric estimates in small student groups.

Purpose of the Study:

• To provide a practical, step-by-step guide for applying bootstrapping to exam questions.
• To demonstrate the utility of bootstrapping for improving the psychometric quality of assessments.
• To facilitate accurate evaluation of exam questions, particularly with limited sample sizes.

Main Methods:

• Utilizing R and SPSS for bootstrapping exam question data.
• Implementing code to resample data and generate robust confidence intervals for key metrics.
• Calculating item difficulty and point-biserial correlations through bootstrapping.

Main Results:

• Bootstrapping provides reliable confidence intervals for psychometric metrics, even with small cohorts.
• The method enables more accurate evaluation of exam question quality compared to traditional parametric approaches.
• Automated visualizations and exportable tables streamline data analysis and presentation.

Conclusions:

• Bootstrapping is a valuable nonparametric technique for enhancing exam question quality assurance.
• This approach improves assessment accuracy in educational settings with small student populations.
• The provided guide and code empower educators to make data-driven decisions for exam revision.