Meeting with Alan today to discuss outline. Came up with basic break-down of paragraphs:
Scope: Previous work, scope simple example existence of warning signal/CSD is clear. Identifiability/Power not clear. Mention the issue of Replicates
Nonlinearity not observed with infinite data
reasons detection fails: Identifiability issue & Power issue. Example signal that exists but inadequate power to detect?
Probability of data vs Probability of a statistic
More knowledge of system help choose the right nonlinear model corresponding to the system.
(Linear/nonlinear estimation analogous to parameterized estimation of tails of distribution vs non-par)
Theoretical mathematics question: Existence Statistical Question: Identifiable, Power
Setting up more exploratory simulations and writing. Simulation speed is quite limiting at the moment. May need to move likelihood calculations into C.
Some discussion of plans for a workshop on NSF data sharing. Alan raises several good points:
Theoretical ecology policy should include a policy on code
How about large data-sets separate from publications, i.e. NEON?
Dave Schimmel NEON – Open Data