The autocorrelation coefficient as a tool for assessing goodness of fit between bioassay predictions and measurement data

Matthew Puncher*, A. Birchall, James Marsh

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

8 Citations (Scopus)

Abstract

Project IDEAS has produced guidelines for internal dose assessment. An integral part of this process is assessing the goodness of fit of biokinetic models to bioassay data. It is recommended that a fit should only be accepted if (a) it is close enough to the data not to be rejected by a χ2 test and (b) if it looks acceptable to 'the eye'. The latter criterion was added to enable the assessor to reject fits which seemed to display some sort of systematic bias. However, there are problems with both of these tests: (a) the χ2 test is dependent on the assumed uncertainties which are often unknown, (b) 'by eye' assessment is subjective. In this paper, another statistic, the autocorrelation coefficient of the residuals, ρ, is investigated. The main advantages of the ρ statistic are that it is objective, very sensitive to biasing and independent of the assumed errors.

Original languageEnglish
Pages (from-to)370-373
Number of pages4
JournalRadiation Protection Dosimetry
Volume127
Issue number1-4
DOIs
Publication statusPublished - 2007

Fingerprint

Dive into the research topics of 'The autocorrelation coefficient as a tool for assessing goodness of fit between bioassay predictions and measurement data'. Together they form a unique fingerprint.

Cite this