Applying Hidden Markov Models in Social Sciences
MetadataShow full item record
The increasing availability of intense longitudinal data in social sciences asks for appropriate analysis methods. This thesis focuses on the use of Hidden Markov Models in social and behavioural sciences. Hidden Markov Models are not widely known/used, which may have to do with unfamiliarity or perceived limitations under researchers. This thesis will aim to give some insights on what can be gained from using a Hidden Markov Model. The analysis in this thesis is done with data about children's behaviour towards other children or robots. A covariate (if the child is interacting with another child or a robot) is taken into account. A Hidden Markov Model and a chi-squared test were both applied on the data and the outcomes were compared. Both the Hidden Markov Model and the chi-squared test revealed differences in behaviour between children interacting with other children or children interacting with a robot. The Hidden Markov Model gave extra information about children's transitions between different types of behaviour, compared to the chi-squared test. There were also differences found between the two conditions on specificc variables. The heterogeneity which is allowed by the Hidden Markov Model also proved to have added value. A few assumptions in preparation of the data were made. This could influence the outcomes of both the Hidden Markov Model and the chi-squared test. For further research it would be interesting to add more covariates in the analysis to reduce the unexplained variance. It would also be interesting to look at opportunities to make Hidden Markov Models more accessible for researchers.