Nice - specific and direct, thanks!

These are quite nice, but can you say a little more about the nature of the uncertainty to help explain to a (non-vis) reader what is being shown and how? E.g. “The transparent coloured bands around these parallel coordinates plots [cite] show 95% confidence intervals (or whatever!) around multiple outputs (or whatever) from a model of …. whatever!” Honestly, - the reader has no idea what this is showing so it’s really hard to engage or be convinced that this is useful. The same is true of the Jones examples - what do they show and how? Sized according to …? Etc!

So much clearer - nice, thanks!

Can you pick a specific thing that you focussed on - what sort of error, what sort of visualization (and why?)

This has potential - it seems as though you have done something useful and interesting, but it’s very hard to understand what you have done from the text. Can you annotate? You say that “Errors from the output can be easily observed in the line chart below” - but I cannot see this> I am not particularly smart, but I do know RAMPVIS, so we should probably be concerned that I cannot see or understand the errors that you claim are easy to observe! Help the reader!

"a first under the epidemiological setting" is important. Do you have a quote? Do try to confirm! The more you can say about use of uncertainty and the user story the better!

OK - but isn't this whole thing about uncertainty, so why can't you account for that in the VIS?

👍😉

This is pretty good but lacks a bit of coherence. Can you relate the reflection and description to the user story? How was uncertainty shown (and why, in light of examples and guidance in the literature). Hope well did this work (were there challenges?) and how did this help the ERMs? Can you explain you examples (6) and those in section 5 more fully - to somebody who has not been working on this project for 6 months! Indeed, to somebody who is not a VIS researcher! Reflection is informative - great!

This works pretty well up until section 6 where it feels like a system description rather than a consideration of error and uncertainty. This may be a problem with trying to deal with error and uncertainty (which, title, 3, 4 & 5 focus on) and model parameters, which cloud the picture and obscure the focus a little. I suggest focusing solely on error and uncertainty, how this is directly depicted and how what we know about doing this works or fails in this context. Parameter sensitivity analysis and model inputs / outputs are dealt with pretty full yin other notebooks - please engage with the team to try to tease these notebooks apart. There is lots of good work and useful experience to report here, but try to revise things a little to differentiate from other notebooks and achieve focus.

I do have something of a concern here in that all seems clear up until now - we are focusing on direct encodings of 3 (?) types of uncertainty : aleatoric/statistical uncertainty, structural uncertainty and prediction uncertainty. But what follows does not evidently do this. Where are the direct error and uncertainty encodings in the four figures? I hope we can be much more explicit about this and focus specifically on the concept that we are are using and assessing here. The same is true of the reflection - see below!

Recommendations and reflection are interesting, but don't seem to me to relate exactly to the concept in hand - representation of error and uncertainty. Of the 9 bullet points - all of which are fine in the broader context - only 1 seems to me to be about error and uncertainty! If I am correct - can we tighten this up a bit? If I am wrong, - can you communicate this more persuasively?

Do you need this one: "interactively compare models with different input, output and configurations" Or does that cloud the picture and obscure the focus? Consider dropping it? Other idioms seem to deal with this.

The title suggests that it should be dropped?