Change sampled parameters

This merge request will allow you to sample in non-standard parameters in a pain-free manner.

This is mostly working, although I may be seeing some bias in the non-standard parameters, e.g., chirp mass.

added 71 commits

• 521a38cf...2835a35e - 70 commits from branch master
• 92df09f8 - Merge branch 'master' into 'change_sampled_parameters', done in the web…

Compare with previous version

added 3 commits

• 07e5b893 - remove some safety featues from waveform generator to make conversion work
• af2a9b03 - add back conversion
• 5cf8227c - add tutorial showing how to sample in chirp mass and mass ratio

Compare with previous version

added 1 commit

• e76c9dd9 - remove more error raising

Compare with previous version

added 5 commits

• f1b6e9d2 - add conversion module
• 998cf6e1 - move conversion functions to new module
• 8bae7d02 - update to use new parameter conversion
• 47980fef - update to use new parameter conversion to make a monolithic posterior
• 2df8bc2c - update syntax in example

Compare with previous version

added 3 commits

• 8092c87e - fix float issues in python 3
• 6d9f681d - update warning
• a8eaff20 - add catch for case where no injection is used, also add print statement after sampler is finished

Compare with previous version

added 2 commits

• 060d47a9 - fix spin conversion in post processing
• 1fb12c67 - typo fix

Compare with previous version

added 8 commits

• 1fb12c67...137a739d - 7 commits from branch master
• 5da5cdf2 - Merge branch 'master' into 'change_sampled_parameters'

Compare with previous version

added 7 commits

• 5da5cdf2...7d1a41de - 4 commits from branch master
• 39a78d1c - Merge branch 'master' into change_sampled_parameters
• fb97356c - Merge branch 'change_sampled_parameters' of git.ligo.org:Monash/tupak into…
• b68a8129 - Merge branch 'master' into change_sampled_parameters

Compare with previous version

added 13 commits

• b68a8129...16041926 - 11 commits from branch master
• 7a49f4ea - Merge branch 'master' into change_sampled_parameters
• e7df84fe - Merge branch 'master' into change_sampled_parameters

Compare with previous version

added 49 commits

• e7df84fe...9cab1ac3 - 41 commits from branch master
• 7c95d798 - Merge branch 'master' into change_sampled_parameters
• 258abaef - tupak/waveform_generator.py
• 93d49f38 - change example script
• 86c979b8 - change format of conversion.py
• 0adf8200 - remove redundant code
• 159c2141 - modify fill_prior and logic for testing overlapping parameters
• 45e78239 - change sampler syntax to allow different parameters
• eff653b1 - remove vestiges of conflict resolution

Compare with previous version

added 11 commits

• eff653b1...a95bf50d - 10 commits from branch master
• 7dd7212b - Merge branch 'master' into 'change_sampled_parameters'

Compare with previous version

unmarked as a Work In Progress

I've removed the WIP label from this.

The behaviour of basic tutorial is unchanged. It can handle a range of extra parameters. Not all of the ones I'd eventually like but critically it can sample in different mass parameters.

The post-processing fills in all the known parameters.

It logs a warning if you specify priors for conjugate parameters, e.g., iota and cos_iota and will sample the default, in this case, iota.

@gregory.ashton can you look over this and merge if you're happy?

added 2 commits

• b8652fbb - add warning if you try to sample from a non prior
• de8faa83 - Merge branch 'change_sampled_parameters' of git.ligo.org:Monash/tupak into…

Compare with previous version

There is a lot of code to go through there, I'm happy to do it though, just now right now. One question, I see that this uses the waveform_generator outside of the likelihood, how important is this? For things like the HierarchicalLikelihood, we won't have a waveform generator. I should get to this tomorrow, but if you are worried about having to maintain changes to master then feel free to merge (since it is passing the tests) and we can fix other things at a later date.

Eric will be happy! :) Nice work

added 5 commits

• de8faa83...a313d514 - 4 commits from branch master
• 84eaf326 - Merge branch 'master' into 'change_sampled_parameters'

Compare with previous version

Good point, I'll see if I can remove the references to the waveform generator from sampler.py, it can probably be done by passing around the likelihood.

For the conversions done during post processing it is necessary to use the waveform generator but these are specific to CBCs so wouldn't be used for hierarchical inference.

added 12 commits

• 84eaf326...82db951c - 4 commits from branch master
• 503da2dc - Merge branch 'master' into change_sampled_parameters
• 8b6da1c7 - move where the non-standard parameters are specified
• 46778b7d - change an attribute name, update docstring
• 6e7a6c13 - add sampling keys attribute
• 286011fd - bbh parameter filling uses likelihood rather than wg and ifos
• 58aaf23f - bug fix
• 15db7783 - change input parameters of samples_to_data_frame
• 3f734cc2 - remove references to wg

Compare with previous version

added 6 commits

• 3f734cc2...469c5ee5 - 5 commits from branch master
• 4df29f1d - Merge branch 'master' into 'change_sampled_parameters'

Compare with previous version

Okay, I'm pretty sure I've removed all references to waveform generator where it shouldn't be referenced.

Can we please merge this so I can stop fixing conflicts?