Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Possible to include calcualtion of Bayes Factor according to Dunstan et al (2022)? #499

Open
MattWenham opened this issue Apr 6, 2023 · 3 comments
Assignees
Labels
enhancement New feature or request

Comments

@MattWenham
Copy link

Dunstan et al in Easy computation of the Bayes factor to fully quantify Occam’s razor in least-squares fitting and to guide actions present "an easy way of calculating [the Bayes Factor] so that it can be routinely used with all least-squares fitting to complement and augment other figures of merit".

Would it be possible to include these calculations as an alternative ic_type?

@Mamba413
Copy link
Collaborator

Mamba413 commented Apr 6, 2023

Does the Bayes factor is equivalent to the Bayesian information criterion? If so, I think it is very direct to implement it.

@MattWenham
Copy link
Author

MattWenham commented Apr 6, 2023

No. From the publication:

…despite its name, the BIC is not Bayesian, and nor are the various related criteria. This is because they do not take into account the prior probabilities of the models. The Bayes factor does.

The basic calculation for the Bayes factor appears to be:

$${\text{MLI}} = \left( {2\pi } \right)^{n/2} L_{max} \frac{{\sqrt {{\text{det }}{\mathbf{Cov}}{p} } }}{{\mathop \prod \nolimits{i = 1}^{n} {\Delta }p_{i} }}$$

…with more explanation in the publication.

@Mamba413
Copy link
Collaborator

Mamba413 commented Jun 6, 2023

I’m sorry for the delayed response. I have been very busy with my doctoral thesis recently.

Today, I carefully read the article on https://www.nature.com/articles/s41598-021-04694-7 and my conclusion is that we will not consider implementing Bayes factor for the time being. The main reason is that it requires the user to input $n$ parameters $\Delta p_i (i=1, \ldots, n)$ where $n$ is the number of variables. And $\Delta p_i (i=1, \ldots, n)$ are indispensable for computing the Bayes factor:
image

As the authors say about the ranges, $\Delta p_i (i=1, \ldots, n)$, have to be decided upon and input by the user. In the context of high-dimensional data analysis, it is not practical or elegant to implement this feature. Therefore, we will not focus on implementing it at this time. We appreciate your suggestion and are always happy to learn from your comments. If you have any further questions or concerns, please feel free to contact us.

@Mamba413 Mamba413 self-assigned this Jul 11, 2023
@Mamba413 Mamba413 added the enhancement New feature or request label Nov 23, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancement New feature or request
Projects
None yet
Development

No branches or pull requests

2 participants