Questions about figure 2

[gold2718]

Why the different color schemes in figure 2?
I expected the right-hand figure to have four red corner squares, four green edge squares, and one blue center square.
By plotting (I think) nine individual probability distributions on the right, I cannot tell if you have really reduced the bias or just smeared everything out so it is impossible to tell. Why the different approach?

Also, I feel this figure would be hard to understand for a color blind individual. For the left-hand figure, different shapes could be used in addition to different colors. Those shapes could also be used to indicate the different biased probability plots. For the right-hand figure, different stipple patterns could be used. This is in keeping with the 'Accessibility' section of the author guidelines (https://www.ametsoc.org/ams/index.cfm/publications/authors/journal-and-bams-authors/figure-information-for-authors/#Accessibility).

[adamrher]

hi steve, I agree that for the right figure, creating the three categories you propose would result in a much less noisy figure. But I don't see how doing it the way I did it, would smooth out any signal. What I intended with the figure is for people to notice that the envelope of all 9 curves, clearly spans a smaller range of probabilities, then the envelope of the three curves on the left figure.

Your point is well taken re: difficulty for color blind individuals. For now I'm going to keep the figure as is. I am open to experimenting with the colors when (and if) the reviews come in.

[gold2718]

But does the figure show that the within-element "bias has been eliminated" as claimed in the caption? Without a similar grouping, I cannot tell. I guess at this point, it is up to a reviewer to be as picky as me.

[adamrher]

ok, I'm curious enough. the paper is submitted but I will keep this issue open, and get back to you.

[adamrher]

Here is a plot where I do as goldy suggests. I think it is an improvement over the previous plot, since it is fair to say that their are 3 types of control volumes that have a different distribution of quadrature nodes within the control volumes (Corners, Edges and the Middle).
temp_pdf.pdf

[adamrher]

and I will change the figure caption to say, the bias has been reduced significantly, as opposed to 'eliminated'

[paullric]

It seems that the phys grid is still not perfectly homogeneous, but there is nonetheless a dramatic improvement over the native SE grid. Is it worth trying to quantify this somehow? ~ Paul

…

On Jul 6, 2018, at 1:00 PM, adamrher ***@***.***> wrote: Here is a plot where I do as goldy suggests. I think it is an improvement over the previous plot, since it is fair to say that their are 3 types of control volumes that have a different distribution of quadrature nodes within the control volumes (Corners, Edges and the Middle). temp_pdf.pdf <https://github.com/PeterHjortLauritzen/2017-MWR-physgrid/files/2171783/temp_pdf.pdf> — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#7 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AFFX5MrVtACfuEJMXzoC67eBRw_4sDxLks5uD8HmgaJpZM4TVr8J>.

[adamrher]

I think the qualitative improvement is the bigger sell. Quantifying the difference between pg2 and pg3, however, might be a worthy exercise ...