What does factor analysis leverage to construct latent factors?

Factor analysis is a statistical technique primarily used to identify and verify the underlying relationships between observed variables. It achieves this by looking at the correlations between these variables. When researchers perform factor analysis, they analyze how variables are related to one another to identify the underlying structures or latent factors that explain these relationships.

The approach focuses on the patterns of correlations—how closely related different variables are to each other—enabling the construction of latent variables that encapsulate the shared variance amongst the observed measures. This is particularly useful in contexts like psychometrics or social sciences, where latent constructs such as intelligence or satisfaction are represented by multiple observable indicators.

In contrast, while survey responses, student feedback, and qualitative interviews might serve as sources of data, they do not directly drive the mechanism of factor analysis itself. Instead, they may provide the raw data from which the correlations are computed, but the essence of factor analysis is its reliance on understanding those correlations to derive constructs. Thus, leveraging correlations between variables is the fundamental aspect of what factor analysis is designed to do.