CEU Electronic Theses and Dissertations, 2015
| Author | Medzihorsky, Juraj |
|---|---|
| Title | Essays on Political Science Applications of the Mixture Index of Fit |
| Summary | This thesis proposes new applications of the Rudas-Clogg–Lindsay mixture index of fit and log-linear models that improve inferences in several areas of substantive research in political science. These include problems from electoral research–detection of electoral fraud from digit distributions, allocations of seats according to votes, territorial variability of electoral support and competition, and analysis of voter transitions with aggregate data–as well as analysis of roll call data in the study of legislative politics, and statistical analysis of political text. The improvements are due to the fact that the methods allow to abandon conventional assumptions known to be difficult or false. Most importantly, the mixture index abandons the assumption that the whole population is described by the model. Furthermore, the index also allows to abandon the assumption that the data was stochastically sampled. Log-linear models allow to represent associations in multivariate categorical data without assuming continuity or requiring transformations that produce it. The thesis is accompanied by an R package named pistar that implements procedures for the application of the mixture index of fit in a variety of settings. |
| Supervisor | Rudas, Tamas |
| Department | Political Science PhD |
| Full text | https://www.etd.ceu.edu/2015/medzihorsky_juraj.pdf |
Visit the CEU Library.
© 2007-2021, Central European University