Development and validation of the professional identity generic scale for chinese university students: Factor-analytic and item response theory approaches

Professional identity can begin to form before entry into full professional practice, yet cross-domain measures for university students must distinguish this emergent identity from identification with an academic major or discipline and must accommodate a strong general factor. A 42-item pool was examined in a construction sample ( n = 836), and the 34-item Professional Identity Generic Scale (PIGS) was evaluated in a validation sample ( n = 1,533). Analyses included oblique exploratory factor analysis (EFA), competing confirmatory factor models, bifactor reliability indices, content-validity indices, sensitivity analyses for uniform response profiles, and a confirmatory bifactor graded response model (B-GRM). The B-GRM evaluation explicitly examined ordered response thresholds/category functioning, item information, local dependence, and differential item functioning (DIF). Parallel analysis suggested three broad factors, whereas theory-guided six-domain EFA retained the intended content blueprint. The bifactor CFA provided the best approximate fit (CFI = 0.942, TLI = 0.934, RMSEA = 0.059, SRMR = 0.034); omegaH = 0.936 and ECV = 0.771 supported a strong general score, while domain-specific omegaHS values were 0.050-0.344. In the B-GRM, general discriminations ranged from 1.81 to 3.72, general-axis thresholds were ordered (-4.30 to 2.04), and the global correlation RMSR was 0.050. Only one adjusted Q3 pair exceeded 0.20 (Items 29–31, 0.207), and no gender or study-stage DIF effect reached the prespecified practical criterion (Delta R2 > = 0.02). Within the sampled Chinese university-student context, the PIGS is best interpreted as a measure of a strong general professional-identity construct sampled through six theoretically specified content domains. The total score is the principal score supported by the evidence; domain summaries may be used descriptively but not as independent latent scales.