Closed fjaviersanchez closed 10 months ago
I tried to run this new version for the Y1 yaml example and I ran into issues towards the end. I copy the log.
Pipeline ran okay. Applying additional prior sigma = 0.2 to params--omega_c This will be assumed to be centered at the parameter center regardless of what the ini file says The limits of the parameter will also not be respected.
Applying additional prior sigma = 0.8 to params--w0 This will be assumed to be centered at the parameter center regardless of what the ini file says The limits of the parameter will also not be respected.
Applying additional prior sigma = 2.0 to params--wa This will be assumed to be centered at the parameter center regardless of what the ini file says The limits of the parameter will also not be respected.
Applying additional prior sigma = 0.3 to params--h This will be assumed to be centered at the parameter center regardless of what the ini file says The limits of the parameter will also not be respected.
Applying additional prior sigma = 0.14 to params--sigma8 This will be assumed to be centered at the parameter center regardless of what the ini file says The limits of the parameter will also not be respected.
Applying additional prior sigma = 0.08 to params--n_s This will be assumed to be centered at the parameter center regardless of what the ini file says The limits of the parameter will also not be respected.
Applying additional prior sigma = 3.9 to params--a0 This will be assumed to be centered at the parameter center regardless of what the ini file says The limits of the parameter will also not be respected.
Applying additional prior sigma = 0.5 to params--bias_lens0 This will be assumed to be centered at the parameter center regardless of what the ini file says The limits of the parameter will also not be respected.
Applying additional prior sigma = 0.5 to params--bias_lens1 This will be assumed to be centered at the parameter center regardless of what the ini file says The limits of the parameter will also not be respected.
Applying additional prior sigma = 0.5 to params--bias_lens2 This will be assumed to be centered at the parameter center regardless of what the ini file says The limits of the parameter will also not be respected.
Applying additional prior sigma = 0.5 to params--bias_lens3 This will be assumed to be centered at the parameter center regardless of what the ini file says The limits of the parameter will also not be respected.
Applying additional prior sigma = 0.5 to params--bias_lens4 This will be assumed to be centered at the parameter center regardless of what the ini file says The limits of the parameter will also not be respected.
Postprocessing...
Generated covariance matrix was not positive definite - beware! /Users/chisa002/Documents/lsst_augur/augur/augur/postprocess.py:151: RuntimeWarning: invalid value encountered in sqrt
heights = 2 np.sqrt(0.5 (C[0, 0] + C[1, 1] - D))
/Users/chisa002/Documents/lsst_augur/augur/augur/postprocess.py:161: RuntimeWarning: invalid value encountered in sqrt
sig0 = np.sqrt(C[0, 0])
Traceback (most recent call last):
File "/Users/chisa002/opt/anaconda3/envs/forecasting/bin/augur", line 7, in
Did you compare our own forecast with one from cosmosis for the same setup?
hmmm, there's something going on with the updates. The matrix is not invertible anymore and for some reason the results don't agree 100% here is the comparison between diagonals (it might be related to the fact that the matrices are not invertible):
I thought it was the priors, but I added the priors in the calculation and they don't seem to make much of a difference. I am comparing here the matrices and they look alike, at least.
cosmosis
numdifftools
This PR is stale now, closing. Will resume the work on a different branch/PR
This PR tries to address #16, implementing our own derivatives of the loglikelihood to get the Fisher matrix. It is still work in progress.