ENBIS: European Network for Business and Industrial Statistics

The Variogram and autocorrelation: Same or different

22 September 2009, 15:05 – 15:25

Abstract

Submitted by

Erik Monness

Authors

Erik Mønness and Shirley Coleman

Affiliation

Hedmark University College, Newcastle University

Abstract

Much of the work of the statistical practitioner results in data that is ordered in some way. Sometimes there is a time order, for example daily output, sometimes there is a spatial order, for example crop yields across a field. The data potentially yield important information about periodic effects, such as diurnal or seasonal time effects or gradual changes in soil type, exposure or spatial usage. Autocorrelation looks at the relationship between successive data values and is widely used in finance, pharmaceuticals and manufacturing. Variograms also look at variation between successive points. Variograms are less widely used, and were originally proposed in the context of soil science and gold mining. The theory behind auto-correlation is designed for a stationary time series. The variogram is believed to give relevant information even if the series is non-stationary. However both statistics can be calculated without making model assumptions. This talk will present, explore and compare the two ways of exploring a one-dimensional series on real data.

View paper

Return to programme