Premium
Poll of Polls: A Compositional Loess Model
Author(s) -
Bergman Jakob,
Holmquist Björn
Publication year - 2014
Publication title -
scandinavian journal of statistics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.359
H-Index - 65
eISSN - 1467-9469
pISSN - 0303-6898
DOI - 10.1111/sjos.12023
Subject(s) - polling , opinion poll , mathematics , sample (material) , set (abstract data type) , order (exchange) , econometrics , statistics , politics , computer science , political science , public opinion , economics , law , thermodynamics , physics , finance , programming language , operating system
A large number of polls on party preferences are published today. In order to obtain an estimate of the changes in political opinion, the polls may be combined into a poll of polls. We discuss a method for combining polls using the fact that they are compositions and respecting the properties of the compositional sample space (the simplex). The method is easily implemented and the estimate may be computed in linear time. We provide an example with Swedish data from year 2007 to 2010. The method also allows us to present the deviations between the estimated compositions and the observed. In the data set, we note, for example, differences between different polling institutes.