Correlates of Imposter Syndrome Among Indonesian Sports Science Students

BACKGROUND:

Imposter syndrome, characterized by a persistent sense of intellectual phoniness and pervasive fear of exposure despite objective evidence of competence, is well documented among undergraduate populations yet remains understudied among non-medical Southeast Asian students, particularly within sports science and related health disciplines.

STUDY DESIGN AND PARTICIPANTS:

This cross-sectional study was conducted from June to August 2025 among sports science undergraduates at State University of Makassar, South Sulawesi, Indonesia, following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. Ethical clearance was granted by the Institutional Biomedical Research Ethics Committee of Hasanuddin University (PROTOKOL-UH25010017), and all procedures were conducted in accordance with the 2013 Declaration of Helsinki. Written informed consent was obtained from every participant, who could withdraw at any time without consequence, and no identifying information was disclosed. The master dataset comprised 438 respondents. Fifteen had genuine missing cumulative GPA values. Analyses not involving GPA, including descriptive, reliability, non-GPA correlation, and non-GPA group-comparison analyses, used N = 438. The GPA correlation, GPA-category comparison, simultaneous ordinary least squares regression, Huber M-estimation, median quantile regression, and hierarchical regression models used the same 423 complete cases under listwise deletion. Eligible students in semesters 3 and 4 were approached during scheduled in-person classes using a mixed paper and online administration procedure, rather than the single supervised classroom administration originally registered with ClinicalTrials.gov. Students could complete a paper-based questionnaire immediately during the class break or within 48 hours; alternatively, they could complete an equivalent online version distributed via the official student communication platform. Two reminder messages were sent at 7-day intervals to non-respondents. No monetary or academic incentives were provided for participation. Data were collected using the Rosenberg Self-Esteem Scale (RSES), the Generalized Anxiety Disorder-7 (GAD-7) scale, and the Clance Impostor Phenomenon Scale (CIPS). The analyses included Spearman's rank correlations, Mann-Whitney U tests, simultaneous ordinary least squares regression, hierarchical multiple linear regression, Huber M-estimation, and median quantile regression at τ = 0.50. The analyses identified adjusted associations rather than causal or predictive relationships, consistent with the cross-sectional design.

MEASURES AND INSTRUMENTS:

Self-esteem, anxiety, and imposter syndrome were assessed using the Rosenberg Self-Esteem Scale (RSES), the Generalized Anxiety Disorder-7 (GAD-7) scale, and the Clance Impostor Phenomenon Scale (CIPS), respectively. The RSES consists of 10 items scored on a four-point Likert scale, with total scores ranging from 10 to 40 and categorized as low (10-20), moderate (21-30), or high (31-40) self-esteem. The GAD-7 consists of seven items scored from 0 to 3, with total scores ranging from 0 to 21 and categorized as minimal (0-4), mild (5-9), moderate (10-14), or severe (≥15) anxiety. The CIPS consists of 20 items scored on a five-point Likert scale, with total scores ranging from 20 to 100 and categorized as none or minimal (≤40), moderate (41-60), frequent (61-80), or intense (>80) imposter syndrome.

STATISTICAL ANALYSIS:

Analyses were performed using IBM SPSS Statistics version 25.0 and R version 4.6.1. Internal consistency was assessed with Cronbach's alpha. Because most continuous variables were non-normally distributed, descriptive data used medians and interquartile ranges. Spearman's rank correlations and Mann-Whitney U tests with rank-biserial correlations were applied for bivariate analyses. GPA-related analyses used 423 complete cases after listwise deletion of 15 observations with missing GPA values. Simultaneous ordinary least-squares regression, hierarchical regression, Huber M-estimation, and median quantile regression were conducted. For the primary OLS regression coefficients, estimates used 5,000 bias-corrected and accelerated (BCa) bootstrap resamples (seed = 123). For Huber M-estimation, P-values were derived from the asymptotic normal approximation to robust test statistics. For median quantile regression (τ = 0.50), P-values were calculated using XY-bootstrap standard errors from 5,000 resamples (seed = 123). Model diagnostics assessed residual normality, homoscedasticity, multicollinearity, residual independence, and influential observations. Statistical significance was set at p < 0.05. Adequacy of the analytic sample was benchmarked using Green's sample-size heuristic (N ≥ 50 + 8k), requiring 130 for ten predictors; the complete-case sample (N = 423) exceeded this benchmark.