Applying Rasch model analysis in the development of the cantonese tone identification test (CANTIT).

Area of Science:

• Psychometrics
• Speech Perception Research
• Linguistics

Background:

• Lexical tone perception is crucial for understanding tonal languages like Cantonese.
• Developing reliable and valid tests for assessing this ability is essential, especially in children.
• The Cantonese Tone Identification Test (CANTIT) was developed to measure this skill.

Purpose of the Study:

• To evaluate the internal structure and psychometric properties of the Cantonese Tone Identification Test (CANTIT).
• To validate both a 75-item (CANTIT-75) and a 30-item short form (CANTIT-30) of the test using Rasch analysis.
• To provide construct validity evidence for CANTIT in measuring lexical tone identification.

Main Methods:

• Rasch analysis was applied to assess the internal structure of the CANTIT.
• A 75-item pool (CANTIT-75) was developed and administered to participants.
• A 30-item short form (CANTIT-30) was derived based on statistical fit, item difficulty, and content evaluation.

Main Results:

• Both CANTIT-75 and CANTIT-30 showed excellent fit statistics (infit/outfit < 2.0).
• Rasch Factor Analysis indicated that the primary Rasch measure explained a substantial portion of the variance (40.1% for CANTIT-75, 38.8% for CANTIT-30).
• Residual components explained minimal variance, suggesting unidimensionality.

Conclusions:

• Rasch analysis is effective for developing psychometrically sound short forms of tests.
• Both the CANTIT-75 and CANTIT-30 exhibit satisfactory internal structure.
• The CANTIT is a valid tool for assessing lexical tone identification ability in Cantonese-speaking children.