Removing linear regression support in Statsample

SciRuby / statsample

A suite for basic and advanced statistics on Ruby.

https://github.com/sciruby/statsample

BSD 3-Clause "New" or "Revised" License

95 stars 29 forks source link

Removing linear regression support in Statsample #53

Open lokeshh opened 8 years ago

lokeshh commented 8 years ago

I think its reasonable to remove linear regression support in Statsample because of following reasons:

Its equivalent which is Normal regression lies in Statsample-GLM
This way all regression models would lie in Statsample-GLM
Statsample-GLM already supports predicting on new data which Statsample doesn't yet
Normal regression in Statsample-GLM works without an intercept providing more flexibility
This will keep regression models management limited to Statsample-GLM and Statsample gem would be responsible for other statistic tools other than regression, hence simplifying the management.

v0dro commented 8 years ago

Works. Go ahead.

lokeshh commented 8 years ago

@v0dro Ok. Well, then it also implies https://github.com/SciRuby/statsample/pull/51 would be no longer needed. Right?

v0dro commented 8 years ago

Are there any particular cases in statsample that are not covered in statsample-glm? If no, then you can safely remove linear regression from statsample and not re-implement the formula language in this gem too (I think that's what you're doing given the commits in #51 - I've been out on vacation so could not follow up on daily activities).

lokeshh commented 8 years ago

Yes you are right. I was implementing formula language in #51. I'm removing it now and will now add a PR to remove multiple regression.

lokeshh commented 8 years ago

Statsample is much faster than Statsample-GLM as these benchmark shows.