Data Mining of Early Day Motions and Multiscale Variance Stabilisation of Count Data

This thesis consists of two parts: an exploration of new measures of backbench opinion in the UK House of Commons, and an exploration of variance stabilising transformations of count data. In the first part, we consider the use of Early Day Motions (EDMs) as a means of gauging opinions of Members of Parliament (MPs) over a range of issues. A much used measure of opinion is that of cohesion; how similar MPs from each political party are to each other. We define a new cohesion measure using the signatories of Early Day Motions and explore this measure over a moving time period for each of the main political parties. We then use Early Day Motions for feature selection. We first identify issues which cause individual parties to be more or less cohesive with one another, before setting out methodology to distinguish which issues cause the major political parties to differ in opinion. We then turn our attention to methods of variance stabilisation of count data. Using data of the number of deaths of coalition forces in Iraq, we demonstrate the good variance stabilisation which the data-driven Haar-Fisz transform possesses. We then modify this transformation so that data with negative counts can be variance stabilised. We show its good performance for simulated data and demonstrate its practical use on the central England temperature data set. Finally, we set about incorporating a transformation parameter into the Haar-Fisz methods, so that through the use of maximum likelihood techniques, the transformation primarily attempts to normalise the data, rather than variance stabilise it.