R Willink 2007 Metrologia 44 105 doi:10.1088/0026-1394/44/2/001
R Willink
Show affiliationsThis paper considers the estimation of a quantity from a set of repeated measurements {xi} by the value of t minimizing ∑i |xi − t|p, as in Pennecchi and Callegaro (2006 Metrologia 43 213–19). The estimate with p = 2 is the familiar sample mean. Assumptions that are relevant when preferring an estimate with p < 2 are identified and examined. Theoretical considerations and results of simulations weaken the case for choosing an estimate with p < 2. The simulation results support the regular use of the sample mean or an estimate that involves the value p = 1.5 when the kurtosis of the dataset is sufficiently large.
06.20.Dk Measurement and error theory
Issue 2 (April 2007)
Received 8 December 2006
Published 12 February 2007
R Willink 2007 Metrologia 44 105
M R James 2005 J. Opt. B: Quantum Semiclass. Opt. 7 S198
Christian Foltin 2002 Nonlinearity 15 2053
Q Jia et al 2009 J. Phys.: Conf. Ser. 190 012157
E Despiau-Pujo and P Chabert 2009 Plasma Sources Sci. Technol. 18 045028
Karl-Erik Thylwe 2005 J. Phys. A: Math. Gen. 38 235
Sigfrido Leschiutta 2005 Metrologia 42 S10
Paula Martin 2005 Phys. Educ. 40 245
gauged
Skyrme models
Y Brihaye and D H Tchrakian 1998 Nonlinearity 11 891
Hajime Tanida et al 2009 J. Phys.: Conf. Ser. 190 012061