Standard deviation σ of |$(f\sigma _{8,i}-\overline{f\sigma _8})/\sigma _i(f\sigma _8)$|⁠, where i is a separate fit for each of the methods. We can see, that for the mock covariance, it is close to 1 (as it is supposed to be when all of the fits share the same covariance.), for fitted covariance it is closing on it, and for jackknife usually takes values >1.4, which shows a much higher degree of deviation from what we assumed to be the truth.