Reading
Gene expression as a continuous phenotypic trait – establishing signal for selection over drift. (Bedford & Hartl, 2009) OU vs BM (or just estimation of selection in the OU model relative to tree length) would have been an obvious way to do this, but the paper tries to make the argument that scaling the data and then comparing
\[\frac{\sigma^2}{2\alpha} \]
is some indication.
Model choice section on warning signals
Trying to figure out whether or not to remove this section (previously Appendix H: Model Choice and Model Adequacy). Second-to-last paragraph mentions this, becomes a wimpy paragraph without it. Run this by Alan.
Model choice works for drake example: supports the transcritical bifurcation over the the saddle-node bifurcation.
Hmm, less convincing: IBM model of saddle node, LTC vs LSN:
Should try LSN-simulation data. Demonstrate robustness of initial estimates?
Null results on stable system
Should we include the stable system in Figures 3 & 4?
Should we include the posterior parameter distributions?
References
- Bedford T and Hartl D (2009). “Optimization of Gene Expression by Natural Selection.” Proceedings of The National Academy of Sciences, 106. ISSN 0027-8424, https://dx.doi.org/10.1073/pnas.0812009106.