search

cdt15 / lingam

Python package for causal discovery based on LiNGAM.
https://sites.google.com/view/sshimizu06/lingam
MIT License
367 stars 56 forks source link

Spport for categorical or binary variables #5

Open fccoelho opened 4 years ago

fccoelho commented 4 years ago

Does the Lingam Algorithm allow the inclusion of categorical and dichotomical variables?

sshimizu2006 commented 4 years ago

Unfortunately, such an algorithm is not available now.

xiangyu-sun-789 commented 3 years ago

Hi professor Shimizu, may I ask why LINGAM does not work with categorical or binary variables? For ICA-LIGNAM, it looks like its identifiability is based on ICA. I found a paper that says ICA works with binary data if the data is sparse. Did I misunderstand something? Thank you very much.

sshimizu2006 commented 3 years ago

The validity of the ICA algorithm implemented in the package was theoretically proved under the assumption that the variables are continuous (and non-Gaussian). So, I replied that no such algorithm is available now (at least in the package). But, it does not mean that it does not allow any other distributions. It might be possible to prove that some algorithm using such a binary ICA works.

xiangyu-sun-789 commented 3 years ago

Hi professor Shimizu, thank you very much for the clarification!

tonyabracadabra commented 1 year ago

Hi professor Shimizu, is categorical variables supported by LIM now? (or that’s binary variables only) @sshimizu2006

sshimizu2006 commented 1 year ago

Yes, LiM allows binary variables.