Title: Handling errors in circular data
Authors: Charles C Taylor - University of Leeds (United Kingdom) [presenting]
Marco Di Marzio - University of Chieti-Pescara (Italy)
Stefania Fensore - University of Chieti-Pescara (Italy)
Abstract: Circular data are often recorded imprecisely. This may be due to instrument error, in which case we can treat the errors as being i.i.d.., and seek to recover the density function of the error-free distribution. Data which include times can also be considered as circular (for example, minutes past the hour), and these are often rounded - to the nearest minute, nearest five minutes, nearest quarter of an hour etc. Although the error-free values may be i.i.d., in this case of digit preference, the errors will no longer have the same distribution. We give examples of data which exhibit these different characteristics and investigate corresponding approaches, including deconvolution, to estimate the underlying density.