mrhedmad/fast-cohen

A fast implementation of a Cohen's D calculator

I'm Luca, and these are my repositories.

I work as a PhD student at the University of Turin, Italy in the Complex Systems for Life Sciences program. I'm a bioinformatician, working with everything genomic and transcriptomic related.

I'm an author and the maintainer for (almost) all packages in my laboratory's organization.

If you want to learn more about me, take a look at my ABOUT page. It contains more information on what I do, as well as my Curriculum Vitae and publication lists.

I really like collaborating with others. If you want to collaborate, check the CONTRIBUTING files in the repository for the project you want to contribute to, or contact me directly!

I administrate bioinfo.ds, a bioniformatics-centered community on Discord.

📨 Contact Me

All my contact information lives in my linktree. I strive to keep it up-to-date!

🛠️ Currently working on

This display is provided by anuraghazra/github-readme-stats.

fast-cohen's Issues

Wrong `n_control` imputation in `cohen` function

fast-cohen/src/main.rs

Line 171 in 912cecc

let n_control: T = NumCast::from(case.len()).unwrap();

should be instead

let n_control: T = NumCast::from(control.len()).unwrap();

Notably, your test_cohen() passed because in that case n_case == n_control....
I suggest implementing a new test with 2 vectors of different lengths.

Time for a fast-cohen 0.1.1 :-D ?

Implement Multithreading

We could make the algorithm go faster by implementing multithreading, even at the cost of having weirdly sorted output lines.

"Cohen's D" should be instead "Cohen's d"

This is a somewhat silly issue, but in the literature this statistic is universally referred to by lowercase d (italicized when possible, i.e., Cohen's d). I would correct a little bit everywhere because I don't want to upset the statisticians (you know they are casesensitive).

Bad (misleading) variable name for the "pooled standard deviation"

In the following lines, pooled_var should be replaced with pooled_std (or something similar), to avoid confusion when reading/debugging the code.

fast-cohen/src/main.rs

Lines 173 to 182 in 912cecc

	let pooled_var = (((n_case - NumCast::from(1).unwrap()) * var(&case)
	+ (n_control - NumCast::from(1).unwrap()) * var(&control))
	/ (n_case + n_control - NumCast::from(2).unwrap()))
	.powf(NumCast::from(0.5).unwrap());
	if pooled_var == NumCast::from(0).unwrap() {
	return NumCast::from(0).unwrap();
	}
	(mean(case).unwrap() - mean(control).unwrap()) / pooled_var

Implement more Effect Size metrics

Other effect size metrics could be useful, for example Hedges' g.

A switch (e.g. --metric hedges) can be added to change what metric is computed by fast-cohen.
Perhaps also a change of names would be appropriate at this point.