search

mjvakili / ccppabc

Approximate Bayesian Computation
MIT License
1 stars 1 forks source link

tex: Referee Report #3 #17

Closed changhoonhahn closed 7 years ago

changhoonhahn commented 7 years ago

is it possible to simulate an extreme data situation where the results from the standard analysis are not consistent with the ABC ones? if so, how realistic it is?

changhoonhahn commented 7 years ago

As @davidwhogg and @mjvakili mentioned, examining galaxy clustering at large (BAO) scales would be one possible way. I'm confused by the wording of this comment ... but that would be a realistic scenario. This comparison would require a different approach than the scope of our paper.

mjvakili commented 7 years ago

@changhoonhahn and I came up withe the following example of extreme data situation:

Consider a dataset which consists of objects with different data qualities and signal-to-noise ratios! Using a standard likelihood function after applying a SNR cut to the data will lead to biases if the selection effects are ignored. This problem can be alleviated by employing ABC with a generative forward model of the data which takes into account cuts in luminosity, SNR, etc.

mjvakili commented 7 years ago

@changhoonhahn Let's not mention large (BAO) scales in the text since they are irrelevant to the scope of our paper! But we should mention it in the response to the referee report.

changhoonhahn commented 7 years ago

@mjvakili : The paragraph you added in Section 3.4 that talks about scenarios where the standard approach and ABC might deviate, is that meant to address this referee comment? Or Ref Report comment 2 (issue #15)? Or both?

mjvakili commented 7 years ago

@changhoonhahn : Those two paragraphs are meant to address this comment. But I think Referee Report 2 #16 also includes this!

mjvakili commented 7 years ago

@changhoonhahn : This is really the subset of Referee Report #3 #17 . Since both have to do with the real data!

changhoonhahn commented 7 years ago

@mjvakili : Can you check out the two paragraphs near the end of Section 3.4. I've heavily edited what you added. I think it goes better with the rest of the Section.

mjvakili commented 7 years ago

@changhoonhahn : Can we reword this sentence? https://github.com/mjvakili/ccppabc/blob/master/ccppabc/text/revision/main.tex#L1014 This is not just about the functional form of the likelihood. Selection effects, SNR cuts, and systematics may change the likelihood but not necessarily the functional form of the likelihood.

changhoonhahn commented 7 years ago

@mjvakili : Done.