IMPROVING LISTENING TEST QUALITY THROUGH STATISTICAL ANALYSIS: A CASE STUDY USING SPSS

Abstract

This study investigates the quality of an English listening comprehension test administered to third-year
students at a Vietnamese public university. The research uses SPSS statistical software to evaluate the test through
descriptive statistics, reliability, and construct validity analyses. Results reveal that while the mean score indicates
overall fair performance, the wide score distribution highlights inconsistencies in student achievement. The
Cronbach’s Alpha coefficient of 0.671 suggests moderate internal consistency, with several items showing low or
negative item-total correlations, indicating potential flaws in the test design. Construct validity is partially
supported through item correlations aligned with theoretical expectations. Based on these findings, the study
proposes specific revisions, including rewording ambiguous items and removing poorly discriminating questions,
to enhance the test’s reliability and validity. By presenting a data-driven approach to test evaluation, this paper
provides practical insights for educators aiming to improve language assessment practices.