Self-efficacy, anxiety, AI utilization, and grammar knowledge among Indonesian and Taiwanese EFL learners
Description
This dataset supports a study on the relationships among self-efficacy, foreign language classroom anxiety, artificial intelligence application utilization, and grammar knowledge among undergraduate EFL learners, with gender examined as a moderator. Participants were 155 fourth-semester students enrolled in English education programmes at three higher education institutions, two in Indonesia and one in Taiwan. Data were collected through an online questionnaire covering self-efficacy, anxiety, and AI utilization, together with a proctored grammar test adapted from the TOEFL structure and written expression section. The dataset contains four measured constructs. Self-efficacy is represented by 11 items adapted from general self-efficacy frameworks. Anxiety is represented by 12 items adapted from the Foreign Language Classroom Anxiety Scale, covering communication apprehension, test anxiety, and fear of negative evaluation. AI utilization is represented by 15 items developed from the Technology Acceptance Model with an added readiness dimension, referring to tools such as ChatGPT, Grammarly, and QuillBot. Grammar knowledge is represented by a summed score across 40 dichotomously scored test items. All questionnaire items used a four-point response format, and reverse-scored items were recoded before analysis. Twenty-six additional columns hold the standardized product indicators used to model the two interaction terms. Also included are two SmartPLS 4 output files: the PLS algorithm results reporting loadings, reliability, average variance extracted, discriminant validity, R-square, f-square, and variance inflation factors; and the bootstrapping results reporting path coefficients, t statistics, p values, and indirect and total effects. No personally identifying information was collected. Ethical clearance was granted by the institutional research ethics committee, and all participants gave informed consent before data collection.
Files
Steps to reproduce
This dataset supports a PLS-SEM analysis of the relationships among self-efficacy, foreign language classroom anxiety, AI application utilization, and grammar knowledge in an EFL context, with gender tested as a moderator. FILES 1. SmartPLS_data_with_interactions.csv - analysis-ready dataset, 155 respondents, 66 variables, no missing values. Gender (1 = male, n = 41; 2 = female, n = 114) SE6-SE23 (11 items, self-efficacy, 4-point scale) AX2-AX23 (12 items, FLCAS-adapted anxiety, 4-point scale) AI1-AI15 (15 items, AI utilization, 4-point scale) GKTOT (summed grammar test score, 0-40 possible, observed 10-36) GxAI_1 to GxAI_15 and GxSE_6 to GxSE_23 (standardized product indicators for the two interaction constructs) 2. Moderation_Product_Indicator.xlsx - SmartPLS 4 output of the PLS algorithm (loadings, reliability, AVE, HTMT, Fornell-Larcker, R-square, f-square, VIF). 3. Moderation_Product_Indicator_after_bootstrap.xlsx - SmartPLS 4 bootstrapping output (path coefficients, t statistics, p values, indirect and total effects). DATA PREPARATION An initial pool of 173 responses was screened. Ten cases showing identical values across all Likert items, including reverse-scored items, and eight cases duplicating an earlier response pattern were removed, leaving 155. Reverse-scored items were recoded before analysis. Interaction terms were formed by standardizing Gender and each indicator of the exogenous construct, then multiplying them (standardized product indicator approach). REPRODUCTION IN SmartPLS 4 (v4.0.9.9) 1. Import the CSV. Set all columns to Metric. Delimiter comma, locale US. 2. Build seven constructs: SE, AX, AI, GK (indicator GKTOT), Gender (indicator Gender), GxAI (15 GxAI_ columns), GxSE (11 GxSE_ columns). All reflective, Mode A. 3. Specify ten paths: SE, AX, Gender and GxSE to AI; SE, AX, AI, Gender, GxAI and GxSE to GK. 4. Run PLS Algorithm: path weighting scheme, standardized results. 5. Run Bootstrapping: 10,000 subsamples, bias-corrected and accelerated confidence intervals, two-tailed, alpha .05. Simple slopes for the Gender x AI interaction were computed at the standardized values of the moderator corresponding to each gender group.