Plots

[shangsiwang]

1. [ ] Update 64 pipelines plot with means, fixed width font
2. [ ] update pairwise distance plot
3. [ ] within/across density plot
4. [ ] decision for each decision
5. [ ] rank reliability plot

[jovo]

for 64 pipelines plot, i vote also the following 2 changes:

1. make dot size proportional to number of scans
2. make color correspond to which dataset, let the code be a 5 letter code, where the first letter is {C,A,H,D} for which atlas
3. i'd make the means maybe black stars or something

repeat for using ranks instead of correlation.

what do you think?

[shangsiwang]

Just update plots fixing 2 and 3. What does 1 mean? Do you think we still need a plot for each decision or just a wilcoxon test?

[jovo]

1 means make the dots larger when the number of scans is larger.

i do still think we need a plot for each decision, and wilcoxon test for them. thanks,

[shangsiwang]

I think we can only see some big dots and small dots which really doesn't provide any insights. And We cannot let the size vary too much since we don't have large space for each column.

Do you want to try other ways to show each decision or just modify the existing violin plot?

[jovo]

did you try to vary the dot size? can i see it? maybe log size?

i thought the violin plot with the permutation distribution "underneath" it was good, no?