Purchase this article with an account.
J.E. Coppens, L. Franssen, T.J. T. P. van den Berg; Monte Carlo Simulation of the Compensation Comparison Method for Retinal Straylight Assessment . Invest. Ophthalmol. Vis. Sci. 2006;47(13):1221.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Recently the psychophysical Compensation Comparison method was developed for routine measurement of retinal straylight. The subject’s responses to a series of two alternative forced choice trials are analyzed using a maximum likelihood (ML) approach assuming some fixed shape for the psychometric function (PF). The ML calculation not only gives an estimation for the straylight value, but also for its accuracy, called "expected standard deviation" (ESD). Repeated measures standard deviation was found to be good (an average of 0.07 log units over 2422 subjects of the GLARE study). Questions addressed with Monte Carlo simulation in the present study were: What is an optimum number and placement of trials? Does ESD from the ML analysis represent true SD? Are there systematic deviations (bias) in straylight value outcome? What happens if the PF of an observer differs from the PF used in the ML analysis?
Responses of an observer were simulated by computer using several potential shapes of the PF as found among the population. Simulations were performed with the sampling scheme that is implemented in the commercially available C–Quant, but also with other sampling schemes. Also a mismatch of the shape of the PF of the observer and the PF used in the ML analysis was simulated.
Standard deviation of measurement outcome is proportional to the reciprocal of the square root of the number of trials. The 25 trial C–Quant sampling scheme resulted in 0.07 log unit standard deviation on average over the population. With the shallowest PFs in the population only a mild increase in standard deviation (0.10 log unit) resulted. For most of the population ESD proved to be a proper or conservative estimator of SD. Mismatch between actual PF and assumed PF did not result in bias.
The Compensation Comparison method is a very robust technique for measuring retinal straylight. Sampling strategy as implemented in the C–Quant is efficient and adequate for most practical purposes.
This PDF is available to Subscribers Only