Expanding your Statistical Universe from Hypothesis Testing to Modelling

Expanding your Statistical Universe from Hypothesis Testing to Modelling
Expanding your Statistical Universe from Hypothesis Testing to Modelling

Principal speaker

Associate Professor Sama Low-Choy

Other speakers

Mari Takashima | Daniela Vasco

Many researchers find themselves in the situation where they need to estimate a proportion of things¾for instance of people, events, organizations, locations¾that have something special about them. In social, environmental & behavioural research examples of such a proportion might include: the percentage of people with an issue or needs regarding health, justice or education; or the proportion of locations suffering from environmental impacts, or where rare or pest species occur. Traditional statistics courses often present novices with just a single approach of null hypothesis significance testing (NHST) that can be calculated "on the back of an envelope": binomial test of proportions, or chi-squared tests. Inferences from NHST are often misinterpreted. Even model-based approaches that provide confidence intervals present computational difficulties; novices are typically only taught asymptotic approximations to Normality that only apply to large samples.

It is rare that NHST is placed within the context of multiple statistical paradigms through methods that are not applicable in all situations, and may only be suitable for confirmatory analysis where there is a clear null hypothesis. In addition NHST can be highly sensitive to sample size, with misinterpretations arising, both for small and large sample sizes. This has led to generations of researchers who are unaware of alternative statistical approaches that may be best suited to a particular applied problem. Inference from a proportion provides a useful and simple context for illustrating the options.

In this workshop we walk you through the appropriate use, and practical ramifications, of using different statistical paradigms to approach a simple problem, involving inference about a probability. We use an interactive exercise designed to engage multiple senses, to help through the rather abstract notions involved.

This session will be interactive and will involve discussion in small groups.

Event categories
Event contact details