Quantile estimators for a nonparametric components of variance situation are proposed consistency and asymptotic normality are proved. Situations with different numbers of measurements for different subjects are considered. Measurements on separate subjects are assumed to be independent, whereas measurements on the same subject have a fixed dependence. The estimators are obtained by inverting weighted empirical distribution functions. An “optimal” estimator and a simple estimator based on within-subject averages are studied. Small-sample properties are studied by simulation as an illustration the estimators are applied to give normal limits for differential light sensitivity of the human eye.