I'm not sure if this is a bug, but the del_q is calculated differently in the Matlab and Fortran versions of bnrh_cdf
Fortran has del_q = 1.0_r8 / (ens_size + 1.0_r8)
|
! For unit normal, find distance from mean to where cdf is 1/(ens_size+1) (del_q_. |
|
! Saved to avoid redundant computation for repeated calls with same ensemble size |
|
del_q = 1.0_r8 / (ens_size + 1.0_r8) |
Matlab has del_q = 1.0 / (ens_size + 1.8)
|
% For unit normal, find distance from mean to where cdf is 1/(ens_size+1) (del_q_. |
|
% Saved to avoid redundant computation for repeated calls with same ensemble size |
|
del_q = 1.0 / (ens_size + 1.8); |
which I suspect is a copy/past error of 1.0r8 but I might be missing something
I'm not sure if this is a bug, but the del_q is calculated differently in the Matlab and Fortran versions of
bnrh_cdfFortran has
del_q = 1.0_r8 / (ens_size + 1.0_r8)DART/assimilation_code/modules/assimilation/bnrh_distribution_mod.f90
Lines 137 to 139 in 890a3be
Matlab has
del_q = 1.0 / (ens_size + 1.8)DART/guide/DART_LAB/matlab/private/bnrh_cdf.m
Lines 66 to 68 in 890a3be
which I suspect is a copy/past error of 1.0r8 but I might be missing something