Hello everyone,
In the context of building the FIM (Fisher Information Matrix) in "w_0 w_a CDM" model, I am trying to find consistent results between the constraints obtained with the Pk from CAMB and the Pk from CLASS.
I am using the following formula to compute each element of the FIM:
Image may be NSFW.
Clik here to view.
For now, I observe rather small differences in the constraints between CAMB and CLASS (in the range of 15–20%), as shown below:
1) CAMB: lmax_scalar=1000
wm±0.005096491692049105
wb±0.0016012556160887369
w0±0.22235200391283078
wa±0.7335948888345178
h±0.004598126058072797
ns±0.012991220492020885
s8±0.022051632948729336
FoM=27.58
2) CLASS: lmax_scalars=2000
wm +/- 0.004918808073212963
wb +/- 0.001696214261682862
w0 +/- 0.1530446306306681
wa +/- 0.2035262848054598
h +/- 0.004695052554558583
ns +/- 0.01239532280750975
s8 +/- 0.017136591385086244
FoM = 36.98
As you can see, I observe a significant difference in the constraints on w0 and wa parameters, and I do not know how to reconcile these two sets of constraints between CAMB and CLASS.
I put in attachment the configuration files params_base.ini and explanatory_base.ini, corresponding to CAMB and CLASS, respectively.
I am also including a short animation ( showing the good agreement for the matter power spectra)
However, as soon as I switch to using the Cℓ and the derivatives of these covariance matrices, I notice a significant difference in the parameters w0 and wa.
One strange thing is that if I set lmax_scalars=200 in CAMB, the FoM drops drastically to FoM=3. This indicates an issue with CAMB because it should be able to compute correctly up to the same value as CLASS (lmax_scalars=2000).
Any help is welcome. I believe my problem stems from poorly tuned parameters, and CLASS is richer than CAMB in terms of all these additional parameters, which can lead to errors.
Best regards
In the context of building the FIM (Fisher Information Matrix) in "w_0 w_a CDM" model, I am trying to find consistent results between the constraints obtained with the Pk from CAMB and the Pk from CLASS.
I am using the following formula to compute each element of the FIM:
Image may be NSFW.
Clik here to view.
For now, I observe rather small differences in the constraints between CAMB and CLASS (in the range of 15–20%), as shown below:
1) CAMB: lmax_scalar=1000
wm±0.005096491692049105
wb±0.0016012556160887369
w0±0.22235200391283078
wa±0.7335948888345178
h±0.004598126058072797
ns±0.012991220492020885
s8±0.022051632948729336
FoM=27.58
2) CLASS: lmax_scalars=2000
wm +/- 0.004918808073212963
wb +/- 0.001696214261682862
w0 +/- 0.1530446306306681
wa +/- 0.2035262848054598
h +/- 0.004695052554558583
ns +/- 0.01239532280750975
s8 +/- 0.017136591385086244
FoM = 36.98
As you can see, I observe a significant difference in the constraints on w0 and wa parameters, and I do not know how to reconcile these two sets of constraints between CAMB and CLASS.
I put in attachment the configuration files params_base.ini and explanatory_base.ini, corresponding to CAMB and CLASS, respectively.
I am also including a short animation ( showing the good agreement for the matter power spectra)
However, as soon as I switch to using the Cℓ and the derivatives of these covariance matrices, I notice a significant difference in the parameters w0 and wa.
One strange thing is that if I set lmax_scalars=200 in CAMB, the FoM drops drastically to FoM=3. This indicates an issue with CAMB because it should be able to compute correctly up to the same value as CLASS (lmax_scalars=2000).
Any help is welcome. I believe my problem stems from poorly tuned parameters, and CLASS is richer than CAMB in terms of all these additional parameters, which can lead to errors.
Best regards
Statistics: Posted by Fabien Dournac — January 29 2025