MIXED UP ABOUT MIXED METHODS? AN OVERVIEW OF MIXED METHODS RESEARCH DESIGNS

We frequently read literature that uses “mixed methods research” in geoscience education, but what does this actually mean? Here, we give an overview of the philosophy underpinning mixed methods research and an introduction to common mixed methods designs in order to help novice researchers choose designs appropriate for their research. Mixed methods research aligns with the pragmatist worldview, in which the mixing of both qualitative (e.g., interviews, focus groups, observation protocols) and quantitative (e.g., Likert surveys, concept inventories) methods is deliberately driven by the research question. The design of the study needs to consider the weighting of qualitative and quantitative data as well as the timing of mixing.

We use examples from our work to illustrate common mixed methods designs (as described by Creswell and Plano Clark, 2010). In a triangulation design, qualitative and quantitative data are equally weighted and independently analyzed in order to converge on a single result. For example, quantitative survey data and qualitative interviews were independently analyzed to address the role of teaching in the recruitment and retention of women in the geosciences. Qualitative to quantitative sequential designs are common in instrument development; for instance, thematic coding of climate skeptic videos was used to generate items for a quantitative skeptic message survey. A quantitative to qualitative sequential design allows a researcher to investigate numerical patterns, as when quantitative data indicated that faculty participants liked a training program, and follow-up focus group interviews probed what specifically participants liked. Qualitative methods can be embedded within a quantitative study, as when interviews are used to clarify student responses to a pre/post attitude survey. Conversely, quantitative methods can be embedded within a qualitative study; for example, survey data on teachers' self-efficacy were embedded within a qualitative characterization of their teaching practices. Finally, qualitative data can be transformed into quantitative data for statistical analysis, such as when counting the percentage of codes that emerged during analysis of expert field sketches.