search

stjude / proteinpaint

Data visualization and analysis framework focused on phenotype-molecular data integration at cohort level.
https://proteinpaint.stjude.org/
Other
16 stars 5 forks source link

Use one Lowess regression line option #490

Closed gavrielm closed 1 year ago

gavrielm commented 1 year ago

Both Loess and Lowess regression lines options are available in scatter plot:

image

We should decide on a single one to use. Seems like these options give pretty different regression lines. Maybe the R regression line is more accurate than the d3 regression line?

xzhou82 commented 1 year ago

please reassign to Airen

On Tue, Aug 29, 2023 at 1:13 PM gavrielm @.***> wrote:

Assigned #490 https://github.com/stjude/proteinpaint/issues/490 to @xzhou82 https://github.com/xzhou82.

— Reply to this email directly, view it on GitHub https://github.com/stjude/proteinpaint/issues/490#event-10224394811, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAMLJJLZDFODYT673OHT7KTXXYWN7ANCNFSM6AAAAAA4DK75FI . You are receiving this because you were assigned.Message ID: @.***>

airenzp commented 1 year ago

I can remove the Loess regression if you want. I just dont understand why to remove choices? They are different regressions. I do think we could name them Loess and Lowess, without adding the R suffix that is just a detail of the implementation.

gavrielm commented 1 year ago

Sorry. I did not realize that they were different regressions. But it does seem like they should give similar results in R (see https://stats.stackexchange.com/questions/161069/difference-between-loess-and-lowess#:~:text=lowess%20and%20loess%20count%20iterations,i.e.%2C%20iterations%3Diter%2B1). It also sounds like Lowess is more standard for scatter plots, but I don't know for sure. We can keep Loess as an option, but before doing that we should compare Loess in d3 with Loess in R.