Article published in:Cognitive Individual Differences in Second Language Processing and Acquisition
Edited by Gisela Granena, Daniel O. Jackson and Yucel Yilmaz
[Bilingual Processing and Acquisition 3] 2016
► pp. 105–128
Analyzing individual differences in second language research
The benefits of mixed effects models
Second language (L2) learners differ on myriad attributes, ranging from cognitive abilities (e.g. auditory perceptual acuity, working memory) to linguistic abilities like L2 proficiency to personality or motivational factors. For decades, scholars have been interested in understanding the impact of these individual differences on L2 outcomes as well as on the learner’s responsiveness to instructional treatments. In contrast to categorical factors like gender, when the attribute of interest varies along a continuum, there are various approaches for analyzing the individual difference variable. Simpler approaches involve reducing the complexity of the analysis by binning or splitting a sample into categorical groups (e.g. by performing a median split) then conducting traditional analyses such as t-tests and ANOVAs. But the concomitant loss of information in the individual difference measure can greatly reduce statistical power, thereby rendering it much more difficult to detect true effects — particularly when the effects are small, as they typically are in L2 research. Moreover, the resulting inferences may no longer map cleanly onto the research questions that originally motivated the research. Mixed effects models — a variant of regression — provide flexibility in the analysis of individual differences that can maintain the observed variability in the data so that scholars can directly test the hypotheses motivating their research. In this chapter, through a series of example models worked out in the R statistical software, I demonstrate the feasibility and ease of using mixed effects models to examine individual differences in L2 research. The examples focus on a scenario requiring the analysis of individual differences while also modeling the main and interactive effects of multiple factors (i.e. experimental design manipulations).
Published online: 23 December 2016
Barr, D., Levy, R., Scheepers, C., & Tily, H.
Bates, D., Kliegl, R., Vasishth, S., & Baayen, H.
(under review). Parsimonious mixed models. Retrieved from http://arxiv.org/abs/1506.04967v1 (16 June 2015).
Bates, D., Maechler, M., Bolker, B., & Walker, S.
(2014) lme4: Linear mixed-effects models using Eigen and S4. R package version 1.1–7. http://CRAN.R-project.org/package=lme4
Gelman, A., & Hill, J.
Gelman, A., & Pardoe, I.
Johnson, P. C. D.
Linck, J. A., & Cunnings, I.
Linck, J. A., Schwieter, J. W., & Sunderman, G.
Mathieu, J. E., Aguinis, H., Culpepper, S. A., & Chen, G.
Nakagawa, S., & Schielzeth, H.
Pedhazur, E. J.
R Core Team
(2014) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org/
Raudenbush, S., & Bryk, A.
Simon, J. R., & Rudell, A. P.
Cited by other publications
Jackson, Daniel O.
Monaghan, Padraic, Simón Ruiz & Patrick Rebuschat
Roberts, Leah & Sarah Ann Liszka
Suzuki, Yuichi, Satoko Yokosawa & David Aline
This list is based on CrossRef data as of 28 november 2020. Please note that it may not be complete. Sources presented here have been supplied by the respective publishers. Any errors therein should be reported to them.