P H C Eilers1, M W Borgdorff. 1. Department of Medical Statistics, Leiden University Medical Centre, Leiden, The Netherlands. p.eilers@lumc.nl
Abstract
SETTING: Human observers generally have a strong tendency to round analog measurements or estimates to 'nice' ending digits, such as 0, 5, or even numbers. This is known as digit preference; it is a well-known phenomenon in frequency distributions of indurations collected in tuberculin surveys. Digit preference can distort estimates of prevalence and other statistical parameters. METHODS: We have developed a statistical model that combines smoothing by penalized likelihood and the transfer of counts from non-preferred to preferred digits, to obtain estimates of: 1) the smooth underlying distribution and 2) the amount of digit preference. RESULTS: To illustrate the validity of the model, it was applied to data from several countries. CONCLUSION: With the proposed model, digit preference can be quantified and frequency distributions of indurations can be corrected for it.
SETTING:Human observers generally have a strong tendency to round analog measurements or estimates to 'nice' ending digits, such as 0, 5, or even numbers. This is known as digit preference; it is a well-known phenomenon in frequency distributions of indurations collected in tuberculin surveys. Digit preference can distort estimates of prevalence and other statistical parameters. METHODS: We have developed a statistical model that combines smoothing by penalized likelihood and the transfer of counts from non-preferred to preferred digits, to obtain estimates of: 1) the smooth underlying distribution and 2) the amount of digit preference. RESULTS: To illustrate the validity of the model, it was applied to data from several countries. CONCLUSION: With the proposed model, digit preference can be quantified and frequency distributions of indurations can be corrected for it.
Authors: Ifedayo M O Adetifa; Abdul Khalie Muhammad; David Jeffries; Simon Donkor; Martien W Borgdorff; Tumani Corrah; Umberto D'Alessandro Journal: PLoS One Date: 2015-10-14 Impact factor: 3.240
Authors: Kennedy Kwasi Addo; Susan van den Hof; Gloria Ivy Mensah; Adukwei Hesse; Christian Bonsu; Kwadwo Ansah Koram; Felix Kwami Afutu; Frank Adae Bonsu Journal: BMC Public Health Date: 2010-01-26 Impact factor: 3.295
Authors: Roque Miramontes; Andrew N Hill; Rachel S Yelk Woodruff; Lauren A Lambert; Thomas R Navin; Kenneth G Castro; Philip A LoBue Journal: PLoS One Date: 2015-11-04 Impact factor: 3.240
Authors: Jim Wright; Mawuli Dzodzomenyo; Nicola A Wardrop; Richard Johnston; Allan Hill; Genevieve Aryeetey; Richard Adanu Journal: Int J Environ Res Public Health Date: 2016-03-09 Impact factor: 3.390