Developing Short Forms of the Fisher Temperament Inventory: A Comparative Study Using Machine Learning Models
Measurement, 2026 (ESCI, Scopus)
- Yayın Türü: Makale / Tam Makale
- Basım Tarihi: 2026
- Doi Numarası: 10.1080/15366367.2026.2714400
- Dergi Adı: Measurement
- Derginin Tarandığı İndeksler: Emerging Sources Citation Index (ESCI), Scopus, Education Abstracts, Educational research abstracts (ERA), ERIC (Education Resources Information Center), Psycinfo, Academic Search Ultimate (EBSCO), Engineering Source (EBSCO)
- Anahtar Kelimeler: Fisher temperament inventory, short form, scale shortening, machine learning, feature selection
- Uşak Üniversitesi Adresli: Evet
Özet
Lengthy self-report inventories increase respondent burden and the risk of careless responding, motivating the development of valid short forms. This study developed and evaluated short forms of the Fisher Temperament Inventory (FTI), a measure of four temperament dimensions, using a machine-learning feature-selection approach. Secondary data from a publicly available archive were analyzed (N = 2,988). For each dimension, items were ranked with two complementary feature-selection methods, and candidate short forms were tested with six predictive models under both categorical and continuous scoring, using an independent holdout set. Compact six-item short forms reproduced the full-form scores with high accuracy across all four dimensions (classification AUC =.94–.97; regression R2 =.83–.88), and short-form to full-form correlations were high (r/ρ =.92–.94). Simple linear models and support vector machines offered the best balance of accuracy and stability, whereas tree-based ensembles performed less consistently. The resulting short forms offer a practical alternative for time-constrained assessment, pending independent cross-cultural validation.