Marcella Corduas

Time series clustering and classification by the autoregressive metric

The statistical properties of the autoregressive (AR) distance between ARIMA processes are investigated. In particular, the asymptotic distribution of the squared AR distance and an approximation which is computationally efficient are derived. Moreover, the problem of time series clustering and classification is discussed and the performance of the AR distance is illustrated by means of some empirical applications.

A class of statistical models for evaluating services and performances

Evaluation can be described as the psychological process which a subject has to perform when a subject is requested to give a determination of merit regarding an item (the attributes of a service, a product or in general, any tangible or intangible object) using a certain ordinal scale. This process is rooted in the subject’s perception of the value/quality/performance of the object under evaluation.

Detecting semiotically-expressed humor in diasporic TV productions

Abstract In this article, we suggest a semiotic approach to the study of visual humorous texts. Our method is based on the multimodal script analysis, which is a useful tool for examining not only verbal texts but also more complex texts, which combine the presence of images and sounds with verbally expressed humor. The resulting framework highlights how some visual comic mechanisms may enhance a different perception of semiotically expressed humor. Moreover, we present a statistical model in order to detect and measure how the resolution of some incongruities may also be determined by specific variables, which help to establish the existence and the strength with which the appreciation of humor varies according to the ethnic group of origin. In particular, the study analyzes the clip ‘Jodhpur Station, 1947’ from a very popular British Asian sketch-show, Goodness Gracious Me (GGM). The sketch shares some similar features with the narrative strategies typical of joke-tellers and is characterized by a complex humorous apparatus depending on different levels of understanding relating to encyclopedic, cross-cultural, and even diasporic knowledge of the world.

Prosodic Markers of Saliency in Humorous Narratives

Much of what we think we know about the performance of humor relies on our intuitions about prosody (e.g., “its all about timing”); however, this has never been empirically tested. Thus, the central question addressed in this article is whether speakers mark punch lines in jokes prosodically and, if so, how. To answer this question, this article unites both the recently emerged research agenda grounding spoken discourse analysis in the precision and verifiability of acoustic analysis and a research agenda within the field of discourse and humor focused on the “performance” of humorous narratives. This article presents an analysis of a relatively simple form: the joke or short humorous narrative. The starting point of this analysis is the folk theory of joke-telling. Through instrumental measurement of pitch, volume, and speech rate, this study shows that, contrary to the folk theory of joke-delivery, punch lines are not delivered significantly louder than the preceding text, but rather at a significantly lower pitch and slower speech rate than the text preceding the punch line. In addition, punch lines are often, but not necessarily, signaled by a laughing voice or a smiling voice and are not preceded by significant pauses. This article concludes that the folk theory of joke-delivery is largely refuted. This study further investigates whether the saliency of punch lines, which would predict higher volume and pitch, is less significant than their final position in the narrative, which, being associated with final position in a paratone, or spoken paragraph, predicts that they will demonstrate lower volume and pitch values. The conclusion is that final positioning trumps the saliency of the punch lines and accounts for the significantly lower pitch and lack of significantly higher volume in punch lines.

The distribution of humour in literary texts is not random: a statistical analysis:

The article presents statistical evidence for the claim that the distribution of humor in Oscar Wildes Lord Arthur Saviles Crime and Douglas Adamss The Hitchikers Guide to the Galaxy is not random and differs significantly between both texts. Using the methodology of the General Theory of Verbal Humor, all the instances of humour in both texts were identified and recorded. The distance between each instance was then calculated and subjected to analysis. The statistical model used to prove the hypotheses is explained in some detail and some hypotheses to explain the findings are presented. The significance of the finding that the distribution of humour in long texts is not random is found to lie in having introduced a new fact in need of explanation through literary theories.

Preliminary estimation of ARFIMA models

In this article we propose a preliminary estimator for the parameters of an ARFIMA(p,d,q) model. The estimation procedure is based on the search of the element in the class of ARFIMA models closest to the estimated ARMA model which best fits the observed time series.

Modelling Correlated Consumer Preferences

The CUB model is a mixture distribution recently proposed in literature for modelling ordinal data. The CUB parameters may be related to explanatory variables describing the raters or the object of evaluation. Although various methodological aspects of this class of models have been investigated, the problem of multivariate ordinal data representation is still open. In this article the Plackett distribution is used in order to construct a bivariate distribution from CUB margins. Furthermore, the model is extended so that the effect of rater characteristics on their stated preferences is included.

Comparing Multistep Ahead Forecasting Functions for Time Series Clustering

The autoregressive metric between ARIMA processes has been originally introduced as the Euclidean distance between the AR weights of the one-step-ahead forecasting functions. This article proposes a novel distance criterion between time series that compares the corresponding multistep ahead forecasting functions and that relies on the direct method for model estimation. The proposed approach is complemented by a strategy for visual exploration and clustering based on the DISTATIS algorithm.