Bayesian inference – a way to combine statistical data and semantic analysis meaningfully

Abstract

This article focuses on presenting the possibilities of Bayesian modelling (Finite Mixture Modelling) in the semantic analysis of statistically modelled data. The probability of a hypothesis in relation to the data available is an important question in inductive reasoning. Bayesian modelling allows the researcher to use many models at a time and provides tools to evaluate the goodness of different models. The researcher should always be aware that there is no such thing as the exact probability of an exact event. This is the reason for using probabilistic models. Each model presents a different perspective on the phenomenon in focus, and the researcher has to choose the most probable model with a view to previous research and the knowledge available.
The idea of Bayesian modelling is illustrated here by presenting two different sets of data, one from craft science research (n=167) and the other (n=63) from educational research (Lindfors, 2007, 2002). The principles of how to build models and how to combine different profiles are described in the light of the research mentioned.
Bayesian modelling is an analysis based on calculating probabilities in relation to a specific set of quantitative data. It is a tool for handling data and interpreting it semantically. The reliability of the analysis arises from an argumentation of which model can be selected from the model space as the basis for an interpretation, and on which arguments.

Keywords: method, sloyd, Bayesian modelling, student teachers

URN:NBN:no-29959