Proof of Unidimensionality in Cognitive Test Instrument for Evaluation Science Learning

Abstract

Dimensional analysis to determine the construct that makes up the instrument as one of the requirements before the analysis is often ignored. Proving dimensions is important because the correct specification of the number of latent dimensions is directly related to construct validity and will determine the appropriate software, approach, and analysis model parameters. Therefore, this research aims to prove the unidimensionality of the Cognitive Test Instrument for Evaluation of Science Learning (CTIESL) through Exploratory Factor Analysis with the Principal Component Analysis Extraction Method. The instrument consisted of 13 essay questions and was responded to by science pre-service teachers who took the Science Learning Evaluation Course (98 people). The results of the analysis prove that the CTIESL has a unidimensional construct based on the Scree Test criteria (detected only one point before the elbow on the scree plot) and the cumulative percent of variance extracted, which shows a satisfactory value (79.232%) so that this instrument is proven only to measure one latent trait, namely the cognitive ability to evaluate the science learning. By fulfilling this unidimensional assumption, the psychometric characteristics of the CTIESL can be explored further with Item Response Theory: 1-Logistic Parameters, 2-Logistic Parameters, or 3-Logistic Parameters, depending on needs.

Keywords

• exploratory factor analysis
• principal component analysis
• unidimensional