New outline
- “An increasing recognition”
- “proof-of-principle” - reliability is unknown
- “Cries wolf”
- ROC, Fig 1
- tau
- method
- data, Fig 2
- analysis, Fig 3
- enough data Fig 4
- disclaimer
- conclusion
Former Outline
- Early Warning matter, no-one does statistics
- We need quantify uncertainty stats
- ROC curve shows this trade-off
- ROC also compares different methods & data sizes (i.e. Fig 4)
- Quantifying “increasing variance” etc — we’ll use tau
- We’ll also do a likelihood-based method
- Introduce the datasets
- ROC curves on the data, Fig 3
- Quantifying data necessary, Fig 4
- Disclaimers: Not completely general
- Conclusion
(Marissa’s comments)
I struggled a bit with the flow in these three paragraphs. I see what you’re trying to do (go from
- the idea of a trade-off, to
- what you’ll do, to
- how your ROC methods illustrates the trade-off, to
- how you’ll apply ROCs),
but I didn’t fully understand the trade-off and why you were talking about it so much in the first paragraph until the ROC illustration described in the second paragraph.
Another approach might be to:
have a paragraph about no one having quantified both reliability and sensitivity and their dependence on data quality, ending with the point that that’s the goal of this manuscript. This paragraph won’t mention anything about trade-offs.
Now introduce trade-offs with fig. 1: start a paragraph with a topic sentence like: “Our tool for quantifying reliability and sensitivity, receiver-operating characteristic (ROC) curves, also illustrates the trade-off between the two.”
- This sets you up to get into the details of (a) what ROC curves are and what they tell you, and then
- the why the trade-off happens and why it matters, using your ROC curves (which the reader now has the background to understand) to illustrate. - MB
Implemented comments, mostly. Still mentions trade-off ahead before presenting ROC curve. Sent draft back to Alan.
To Do
- Figure revisions
- Appendix updates
- Caption revisions
- Flag points in supplementary text