CliMA / ClimaCoupler.jl

ClimaCoupler: bringing atmosphere, land, and ocean together
Apache License 2.0
21 stars 3 forks source link

Add "triangle" plot #831

Closed Sbozzolo closed 4 weeks ago

Sbozzolo commented 1 month ago

Let's make it square. :)

In this PR, I add a plot that compares our model with the best model (where best is the one that minimizes RMSEs). Errors are reported in terms of median RMSE

image

Sbozzolo commented 1 month ago

My understanding of the spirit of the triangle plot is to visually compare various models, so I thought that having the best model was important.

As for the triangles, what I can easily do is make them squares. That is, condense all the squares I have now as sub-squares of one square (instead of sub triangles of one square).

Making triangles would add more more code.

On Mon, Jun 3, 2024, 6:55 AM Zhaoyi Shen @.***> wrote:

@.**** approved this pull request.

Looks good to me, thanks! Maybe we can get rid of the best model to make the plot less busy? Would it be hard to do the actual triangle plot (we don't need to do it now)?

— Reply to this email directly, view it on GitHub https://github.com/CliMA/ClimaCoupler.jl/pull/831#pullrequestreview-2093952775, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACF6E7OTZCNUDEJNNC2FOC3ZFRYW5AVCNFSM6AAAAABITTV4S6VHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZDAOJTHE2TENZXGU . You are receiving this because you authored the thread.Message ID: @.***>

szy21 commented 1 month ago

My understanding of the spirit of the triangle plot is to visually compare various models, so I thought that having the best model was important. As for the triangles, what I can easily do is make them squares. That is, condense all the squares I have now as sub-squares of one square (instead of sub triangles of one square). Making triangles would add more more code. On Mon, Jun 3, 2024, 6:55 AM Zhaoyi Shen @.> wrote: @*.*** approved this pull request. Looks good to me, thanks! Maybe we can get rid of the best model to make the plot less busy? Would it be hard to do the actual triangle plot (we don't need to do it now)? — Reply to this email directly, view it on GitHub <#831 (review)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACF6E7OTZCNUDEJNNC2FOC3ZFRYW5AVCNFSM6AAAAABITTV4S6VHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZDAOJTHE2TENZXGU . You are receiving this because you authored the thread.Message ID: @.>

Sounds good, let's keep the best model. And let's move to sub-squares once we have more variables (unless it's very easy to do).