ABSTRACT
The purpose of the present investigation is to discuss current and traditional methods of statistical analysis in sociolinguistics and to show that decision making should be based on the linguistic analysis as a first priority. The results reveal that there is consistency across the factors and conditioning both phenomena analyzed regardless of the statistical models we used. It is argued that this consistency is due to the large number of speakers in the corpus used. The mixed-effects models showed similar results after including random effects in the two examples analyzed.
