Principal Component Analysis vs Lex Jansen（主成分分析与Lex詹森）.pdfVIP

Principal Component Analysis vs. Exploratory Factor Analysis Diana D. Suhr, Ph.D. University of Northern Colorado Abstract Principal Component Analysis (PCA) and Exploratory Factor Analysis (EFA) are both variable reduction techniques and sometimes mistaken as the same statistical method. However, there are distinct differences between PCA and EFA. Similarities and differences between PCA and EFA will be examined. Examples of PCA and EFA with PRINCOMP and FACTOR will be illustrated and discussed. Introduction You want to run a regression analysis with the data you’ve collected. However, the measured (observed) variables are highly correlated. There are several choices use some of the measured variables in the regression analysis (explain less variance) create composite scores by summing measured variables (explain less variance) create principal component scores (explain more variance). The choice seems simple. Create principal component scores, uncorrelated linear combinations of weighted observed variables, and explain a maximal amount of variance in the data. What if you think there are underlying latent constructs in the data? Latent constructs cannot be directly measured influence responses on measured variables include unreliability due to measurement error. Observed (measured) variables could be linear combinations of the underlying factors (estimated underlying latent constructs and unique factors). EFA describes the factor structure of your data. Definitions An observed variable can be measured directly, is sometimes called a measured variable or an indicator or a manifest variable. A principal component is a linear combination of weighted observed variables. Principal components are uncorrelated and orthogonal. A latent construct can be measured indirectly by determining its influence to responses on measured vari