Back to search

University of Southampton

Modelling compositional time series from repeated surveys

Abstract

dc:description.abstract

<p>A compositional time series is defined as a multiple time series in which each of the series has values bounded between zero and one and, moreover, the sum of the series equals one at each time point. Data with such characteristics are observed in repeated surveys when a survey variable has a multinomial response and the interest lies on the proportion of units classified in each of its categories. In this case the survey estimates are proportions of a whole subject to a unity-sum constraint.</p><p>This thesis proposes state-space models for improving estimation of compositional data from repeated surveys taking into account the sampling errors. The proposed modelling procedure provides bounded predictions and signal estimates for the compositions, satisfying the unity-sum constraint, while taking into account the sampling errors. This is accomplished by mapping the compositions from the Simplex onto the Real space using the additive logratio transformation, then modelling the transformed data via multivariate state-space models, and finally applying the additive logistic transformation to obtain estimates in the original scale. In addition it is shown that the modelling procedure is permutation invariant. </p><p>The method is applied to compositional data from the Brazilian Labour Force Survey. The model for the survey estimates is a combination of the multivariate models specified for the signal and noise processes. Estimates for the vector of proportions of labour market status and the unemployment rate are obtained. Estimates of seasonally adjusted series are also produced. The results of the empirical work lead to the conclusion that smoother trends are obtained with a model which explicitly accounts for the sampling errors, when compared with the results from other standard procedures for seasonal adjustment.</p>

Degree

thesis:*
Name dc:type.qualificationname
Ph.D.
Level dc:type.qualificationlevel
doctoral
Grantor dc:publisher.institution
University of Southampton
Year dc:date.issued
1996

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Nascimento Silva, Denise Britz do

Rights

Language dc:language
English, eng

Chain of custody

source
Harvested from
University of Southampton
Base URL
eprints.soton.ac.uk/cgi/oai2
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
related terms
citation

Nascimento Silva, Denise Britz do. Modelling compositional time series from repeated surveys. doctoral thesis, University of Southampton, 1996.