Research Matters: Pretty pictures and parameters – understanding statistics using graphics (Research Counts)
Dr. Graeme Hutcheson
Wednesday 11th June 2014, 12-1pm
Room AG3/4, Ellen Wilkinson Building
Using statistics effectively in education involves applying a range of analyses to model many different types of data; linear models for variables measured on a continuous scale (eg. OLS regression and ANOVA), logit models for categorical variables (eg. logistic, proportional-odds and multinomial) and models for count variables (eg. Poisson regression, chi-square and log-linear).
These models present a challenge for researchers and a steep learning curve for PhD students who often have little or no background in statistics and as a consequence are often daunted by the time and resources required to become statistically literate. This session will argue that it is possible for a comprehensive range of techniques to be learned within the time available to PhD students if the techniques are applied as part of a coherent underlying theory (generalized linear models) and a systematic method of representating these models is adopted (a graphical representation based on predictions).
Although this presentation deals with a range of statistical models, the emphasis is on how to teach a comprehensive system of analysis in a time-frame typically available to a PhD student.