Dependence and Interaction in Frequency Data Analysis in Slaresearch

In SLA research, dependence in frequency data is a prevalent problem (Hatch & Lazaraton, 1991). Researchers usually regard the data as being independent and subject them to statistical analyses. Another problem in frequency data analysis in SLA research is the exclusion of interaction terms. This is the case because of the use of chi-square analysis or the unpopularity of multifactorial frequency data analyses. This study investigates the violation of the “independent observation” assumption as well as the effect of including interaction terms in frequency data analysis. Reanalyzing a couple of published data sets, the paper argues in favor of using multifactorial frequency data analyses over multiple chi-squares in order to take into account dependence and interaction of the frequency data. A set of recommendations for SLA researchers will be provided when statistically analyzing frequency data in SLA research.