ABSTRACT
When researchers, including linguists, use linear models to analyze their data, they often need to “transform” the variables in their study. This chapter discusses linear and nonlinear transformations. Whereas linear transformations have rather innocuous effects (they leave the relationships between data points untouched), nonlinear transformations can substantially alter one’s conclusions (the relationships between data points is affected). The chapter discusses “centering” (subtracting the mean) and “standardizing” (dividing by the standard deviation) as two common linear transformations. The main nonlinear transformation discussed here is the logarithm, which is ubiquitous in linguistics. Extensive examples tell the reader how to interpret linear models when variables have been transformed. The chapter concludes with a discussion of correlation as it can be thought of as a standardized form of regression.
