A Comparison of Principal Components Analysis and Factor Analysis for Uncovering the Early Development Instrument (EDI) Domains

Principal Components Analysis (PCA) and Factor Analysis (FA) are often employed in identifying structures that underlie complex psychometric tools. Although the two strategies differ in terms of their applications, it is important to compare structures that may emerge when they are performed on such tools as the Early Development Instrument (EDI). The purpose of such an analysis is to simplify reported findings by using a reduced set of correlated EDI measurements. We compared the underlying components and factors based on different extraction and rotation methods on EDI data from Alberta, Canada, using a two-part strategy: to report on the component and factor structures without imposing any restrictions on the number of components and factors, and then to report on multiple tests to arrive at a clean structure by retaining only a restricted number of factors. Regardless of the chosen method of extraction and rotation, some items were found redundant in both PCA and FA. The analysis revealed that PCA summarized the structure better than FA (ML), eliminating some redundancy in the number of items while retaining a comparatively better overall variance. The results indicate that items that load on more than one component or factor substantially decrease the ability of PCA and FA to detect an underlying construct, and dropping such items could reduce the amount of complexity in EDI when formulating and testing an explanatory model of child development, especially at a community level. The paper concluded that an important task in analyzing the well-regarded EDI domains involves the identification of items that do not contribute to our understanding of child development, either theoretically or methodologically.