ENBIS: European Network for Business and Industrial Statistics

Multivariate Control Charts for Detection of Outbreaks

22 September 2009, 11:40 – 12:10

Abstract

Submitted by

Linus Schiöler

Authors

Linus Schiöler Kjell Pettersson Eva Andersson Marianne Frisén

Affiliation

Statistical Research Unit, University of Gothenburg

Abstract

Outbreaks in the meaning of a monotonic increase from a constant in-control level can appear in finance, industrial production, public health and other areas. On-line detection of outbreaks of epidemics has recently drawn much interest. A new semiparametric method based on an estimation method by Frisén, Andersson and Pettersson (2009) was proposed in Frisén, Andersson and Schiöler (2009) and was applied to Swedish influenza data. Information about the spatiotemporal patterns of the spread may be utilized to improve the performance. In Schiöler (2008) a time-lag between the metropolitan areas and the local areas was found.
Methods for multivariate surveillance were applied, combining data from the two areas in an optimal way. Since multivariate on-line surveillance problems are complex, it is difficult to find optimal methods. By using the sufficiency principle, optimal reduction of data was found.
Since measures for evaluation of univariate surveillance are not suitable in multivariate sur-veillance, some new measures for evaluation were used. The method was applied to Swedish data on influenza.


References

[1] [Frisén, M., Andersson, E. and Pettersson, K. (2009). Semiparametric Estimation of Outbreak Regression. Statistics, in press.
[2] [Frisén, M., Andersson, E. and Schiöler, L. (2009). Robust Outbreak Surveillance of Epidemics in Sweden. Statistics in Medicine, Vol. 28, 476-493.
[3] Schiöler, L. (2008). Explorative Analyzis of Spatial Patterns of Influenza Incidences in Sweden 1999-2008. Research Report 2008:5, Statistical Research Unit, Department of Economics, Göteborg University, Sweden,

Return to programme