Premium
Bias and precision of letter chart threshold estimates: a simulation study
Author(s) -
Gilchrist Jim
Publication year - 1996
Publication title -
ophthalmic and physiological optics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.147
H-Index - 66
eISSN - 1475-1313
pISSN - 0275-5408
DOI - 10.1046/j.1475-1313.1996.9683359x.x
Subject(s) - chart , statistics , range (aeronautics) , computer science , psychometric function , mathematics , algorithm , psychophysics , psychology , materials science , neuroscience , perception , composite material
Purpose: An ideal letter chart should give accurate (precise and unbiased) estimates of true threshold over the whole range of the chart. Precision and bias may be affected by chart design, scoring procedure and patient response characteristics. Methods: Accuracy of acuity estimates on Bailey‐Lovie (BL) and Snellen (SN) charts were evaluated by computer simulation. Subjects' psychometric acuity functions were modelled using a Weibull function with constant slope and variable threshold level. The model subject gave 500 simulated ‘readings’ of each chart with true thresholds covering the full range of each chart. For each reading a number of scoring procedures were employed based on combinations of: (a) stopping criterion; and (b) estimation rule. Results: Results show that: (1) the BL chart is not necessarily more accurate than SN, but its precision and bias are much more consistent across the threshold range. (2) Accuracy of both charts depends critically on scoring procedure and is affected more by choice of stopping criterion than by choice of estimation rule. Conclusion: A recommended scoring procedure for the Bailey‐Lovie chart based on simulations is: stop reading when 3 errors occur in one line, and estimate threshold as the last letter read correctly.