Change-point estimation in a multinomial sequence and homogeneity of literary style

Abstract

To help settle the debate around the authorship of Tirant lo Blanc, all the words in each chapter of that book are categorized according to their length and the appearance of various words is counted. The graphical exploration of the sequences of multinomial observations obtained reveals a clear single sudden change point that is consistently estimated to be between chapters 371 and 382 and might indicate a switch of author. Correspondence analysis indicates that at the end of the book the words tend to be longer and the frequency of various words changes significantly. By doing a cluster analysis of the multinomial observations, the evidence in favor of the existence of that stylistic boundary is strengthened, because the two clusters obtained match very closely the before and after change-point groups; only a few chapters at the end of the book appear to be misclassified by the change point.