Abstract
This study addresses the statistical analysis of a phenomenon in Russian verbal paradigms, a suffix shift that is spreading through the paradigm and making it more regular. A problem that arises in the analysis of data collected from the Russian National Corpus is that counts documenting this phenomenon are based on repeated observations of the same verbs and, moreover, on counts for different parts of the paradigms of these same verbs. Unsurprisingly, individual verbs display consistent (although variable) behavior with respect to the suffix shift. The non-independence of the elementary observations in our data has to be taken into account in the statistical evaluation of the patterns in the data. We show how mixed-effects modeling can be used to do this in a principled way, and that it is also necessary to do so in order to avoid anti-conservative evaluation of significance.
© 2010 Walter de Gruyter GmbH & Co. KG, Berlin/New York