Elementary teacher students’ science-literacy confidence climbs with academic level, Rasch analysis shows

Self-reported confidence in scientific-literacy skills is not the same thing as objectively demonstrated competence, and a new cross-sectional study from Indonesia treats that distinction as the starting point of its analysis. Researchers at Universitas Pendidikan Indonesia, Universitas Negeri Padang and Universitas Negeri Semarang measured how sure elementary teacher-education students and practicing teachers feel about using scientific-literacy competencies, and how those confidence ratings behave statistically across academic levels. The work, accepted for publication in Frontiers in Education, applied a Rasch rating scale analysis to a 40-item instrument completed by 500 participants at the Universitas Pendidikan Indonesia Tasikmalaya Campus, with 100 participants each drawn from Years 1 through 4 of the undergraduate program and from postgraduate study. The undergraduate participants were pre-service teachers in training, while the postgraduate participants were practicing teachers already working in classrooms.

The instrument asked respondents to rate their confidence on five ordered categories, from 1 for very unsure to 5 for very sure, across four scientific-literacy competency areas. The team analysed the ratings with the Andrich Rating Scale Model using WINSTEPS version 5.7.3.0, a standard approach for converting ordered confidence categories into interval-level person measures expressed in logits. The calibration produced a mean person measure of 2.46 logits, a person reliability of 0.97, and a real person separation of 5.37. Those separation and reliability figures indicate that the instrument distinguished confidence levels among respondents with considerable precision, ordering people along the measured construct rather than lumping them into a few coarse bands.

Item-level diagnostics were broadly acceptable but not flawless. Infit mean-square values ranged from 0.64 to 1.32 and Outfit mean-square values from 0.56 to 1.38, meaning no item distorted the measurement structure enough to warrant removal, although some items fit less cleanly than the ideal value of 1.0. Rasch measures explained approximately 52.9 percent of the raw variance in the ratings. However, the first residual contrast reached 2.78 eigenvalue units, a signal that a secondary dimension remained in the residual variance after the main confidence measure was extracted. Local dependence among items was modest, with a maximum adjusted Q3 statistic of 0.36, and person-item targeting was weak at the upper end, indicating the item pool did not fully cover the confidence range of the most confident respondents.

One of the study’s clearest findings concerns differential item functioning, or DIF, which tests whether an item behaves differently for different groups after the overall confidence level is accounted for. Six items showed DIF associated with academic level, meaning those confidence items did not measure identically across the undergraduate years and postgraduate study. Because the undergraduate respondents were pre-service teachers and the postgraduate respondents were practicing teachers, the authors interpret such DIF as a property of the instrument that should be examined and refined before strong comparisons between these groups are made on those items. The study reports that the results support refinement of item targeting, the residual structure, and the DIF-flagged items before more confident interpretations are drawn from them.

Across the whole sample, continuous person measures differed significantly by academic level. A Welch analysis of variance yielded F(4, 244.79) = 341.15, p < .001, with an omega-squared effect size of 0.623, a very large proportion of variance in confidence measures associated with academic level. The pattern of means climbed across the undergraduate years, but the Year 4 and postgraduate groups did not differ significantly from each other. In other words, measured confidence rose steadily through the four years of elementary teacher education and then plateaued among practicing teachers in postgraduate study, rather than continuing upward once participants entered the profession.

The authors are careful about what these numbers can and cannot support. Because the design is cross-sectional, the data capture confidence at one point in time rather than tracking the same individuals as they progress, so the level differences describe associations with academic level, not demonstrated growth within persons. Because the instrument measures self-reported confidence, the person measures describe how sure respondents feel about their scientific-literacy competencies, not an objective test of whether those competencies are actually demonstrated. The study’s stated purpose is to assess confidence in scientific-literacy competencies across academic levels and to show that confidence should be measured and interpreted as distinct from objectively demonstrated competence.

The practical upshot for teacher education is a measurement agenda. The authors conclude that the findings support refinement of the instrument’s item targeting, attention to the secondary residual structure indicated by the 2.78-eigenvalue contrast, review of the six DIF items, and continued monitoring of local dependence before stronger claims are made about confidence differences between pre-service and practicing teachers. With those refinements in place, the same Rasch framework could yield more precise comparisons of how confidence develops from the first year of elementary teacher education through postgraduate practice, and how that measured confidence relates to competencies demonstrated on objective assessments.

For a field increasingly interested in how elementary teachers prepare to teach science, the study offers a technically grounded template for measuring a construct that is easy to conflate with competence itself. The combination of high person reliability, strong separation, and a large association with academic level shows that confidence in scientific-literacy competencies can be quantified meaningfully with a rating scale approach. At the same time, the residual dimension, the weak upper-end targeting, and the six academically functioning items show exactly where the measurement needs work. The study, received by the journal on 24 August 2026 and accepted on 2 October 2026, positions confidence as a measurable outcome of elementary teacher education in its own right, one that rises with academic level and plateaus among practicing teachers, and one that must be kept analytically separate from the competencies it describes.

Subject of Research: Assessing confidence in scientific-literacy competencies across academic levels in elementary teacher education: a Rasch rating scale analysis

Article Title: Assessing confidence in scientific-literacy competencies across academic levels in elementary teacher education: a Rasch rating scale analysis

Article References: Hamdu, G., Yulianto, A., Fitria, Y., & Susongko, P. (2026). Assessing confidence in scientific-literacy competencies across academic levels in elementary teacher education: a Rasch rating scale analysis. Frontiers in Education, 11, Article 1976534. https://doi.org/10.3389/feduc.2026.1976534

Image Credits: AI Generated

DOI: 10.3389/feduc.2026.1976534

Keywords: Assessing, confidence, scientific-literacy, competencies, across, academic, levels, elementary, teacher, education, Rasch, rating