Note (verbatim, from paper/manuscript_marine.qmd around line 277, under ## Estimator family {#sec:methods-fig}, right after Fig. 1 demo_1_methods.png):
At around what level of dv/v do the methods start to disagree? what overall
differences bias and variance do we expect from these different methods?
Context: the preceding paragraph (lines 259-264) qualitatively states that all seven NoisePy estimators agree at small dv/v, but phase methods (MWCS) cycle-skip and warping methods (DTW/WTDTW) under-shoot at large dv/v — with "no estimator is simply right." The note asks for that to be made quantitative.
Acceptance criterion: add a quantitative statement (in text, and/or a supplementary panel/table) giving the dv/v value at which estimator families diverge, and the bias/variance of each of the seven estimators relative to the known synthetic ground truth across a swept range of dv/v — generated from the existing seven-estimator sweep used for demo_1_methods.png, not a qualitative restatement.
Note (verbatim, from
paper/manuscript_marine.qmdaround line 277, under## Estimator family {#sec:methods-fig}, right after Fig. 1demo_1_methods.png):Context: the preceding paragraph (lines 259-264) qualitatively states that all seven NoisePy estimators agree at small dv/v, but phase methods (MWCS) cycle-skip and warping methods (DTW/WTDTW) under-shoot at large dv/v — with "no estimator is simply right." The note asks for that to be made quantitative.
Acceptance criterion: add a quantitative statement (in text, and/or a supplementary panel/table) giving the dv/v value at which estimator families diverge, and the bias/variance of each of the seven estimators relative to the known synthetic ground truth across a swept range of dv/v — generated from the existing seven-estimator sweep used for
demo_1_methods.png, not a qualitative restatement.