Confidence intervals on fit parameters derived from optical reflectance spectroscopy measurements

Abstract

We validate a simple method for determining the confidence intervals on fitted parameters derived from modeling optical reflectance spectroscopy measurements using synthetic datasets. The method estimates the parameter confidence intervals as the square roots of the diagonal elements of the covariance matrix, obtained by multiplying the inverse of the second derivative matrix of chi2 with respect to its free parameters by chi2/v, with v the number of degrees of freedom. We show that this method yields correct confidence intervals as long as the model used to describe the data is correct. Imperfections in the fitting model introduces a bias in the fitted parameters that greatly exceeds the estimated confidence intervals. We investigate the use of various methods to identify and subsequently minimize the bias in the fitted parameters associated with incorrect modeling.

Diagnosis of breast cancer using diffuse optical spectroscopy from 500 to 1600 nm: comparison of classification methods

“…For each estimated value, a confidence interval computed from the covariance matrix was used to assess the reliability for each fit parameter. 33 …”

Estimation of lipid and water concentrations in scattering media with diffuse optical spectroscopy from 900 to 1600 nm

“…21 The model holds when the reduced scattering is greater than the absorption and the distance between the optical fibers is greater than the scattering length. Moreover, the medium should be homogenous.…”

Are quantitative attenuation measurements of blood by optical coherence tomography feasible?

“…They are derived from the covariance matrix reported by the fitting algorithm and quantify the range in which each fit parameter can be varied without influencing the overall fit statistics. They should be interpreted with caution [11]. A fit using an inappropriate model can yield very small c.i.…”