The mean differences in ln(P_{dc }) between the two levels of each categorical variable estimated by fitting a two factor fixed effects models to the fellow eye differences in ln(P_{dc }). p Values were estimated using two sided t tests