Towards Full (er) Integration in Mixed Methods Research: The Role of Canonical Correlation Analysis for Integrating Quantitative and Qualitative Data

Abstract

One of the biggest developments in mixed methods research has been the conceptualization of one or more analysis types associated with one tradition (e.g., qualitative analysis) being used to analyze data associated with a different tradition (e.g., quantitative data)—what Onwuegbuzie and Combs (2010) called crossover mixed analyses, or, more simply, crossover analyses. A hallmark of crossover analyses is the notion of quantitizing, which, in its simplest form, involves converting qualitative data into numerical forms that can be analyzed statistically. The focus on quantitizing has been on descriptive-based quantitizing approaches such as counting the occurrence of emergent themes. Unfortunately, scant guidance exists on inferential-based quantitizing, which refers to the quantitizing of qualitative data for the purpose of prediction or estimation (Onwuegbuzie, in press). Although recent literature has emerged on a few inferential-based quantitizing approaches (i.e., multiple linear regression analysis, structural equation modeling, hierarchical linear modeling), there still remains some general linear model analyses for which mixed methods researchers, in pursuit of conducting crossover analyses, can benefit from guidelines. One such analysis is canonical correlation analysis. Its importance stems from the fact that the analysis of qualitative data typically yields multiple patterns of meaning (e.g., codes, themes), which then can be correlated with other available variables (e.g., demographic variables, personality variables, affective variables) via the use of canonical correlation analysis. Therefore, the purpose of this article is (a) to describe canonical correlation analysis and (b) to illustrate how canonical correlation analyses can serve as an inferential-based quantitizing using a heuristic example.