ABSTRACT
Significance tests may be considered as an assessment combining three elements: a statistical model of variation, stating, for example, that observations are expected to be Normally distributed about a population mean with standard deviation; an observed effect size, for example, the difference; and the weight of evidence supporting the observation, typically based on the sample size. In this chapter, the author focuses on the second aspect: effect size. Shorn of test machinery, they are descriptive measures of samples. They can be useful for visualisation and citation purposes. Statisticians have occasionally proposed effect size measures alongside tests as a kind of ‘descriptive analogue’ of their test. Bowie et al. discuss whether the present perfect construction is more properly considered a type of ‘present’ or a type of ‘past’ referring construction. Linguistic variables may often be measured as Binomial proportions, expressing the probability that in a random case drawn from a sample the people might find a particular linguistic phenomenon.
