Notes

Nonparametric-bayes

• Establishing chunck dependencies, simplifying and abstracting code.

• still need to fix up Allen MCMC convergence when underlying model is Myers 10:18 am 2013/05/22

• replicate runs on data from allen and myers model 08:39 pm 2013/05/21

Clean code for complicated scripts

Pseudo-code somewhat captured by the chuncks. Ideal would be:

• Define models
• Define parameters
• Simulate data
• Estimate MLE
• Estimate under GPP
• Allen MCMC
• Ricker MCMC
• Myers MCMC
• Determine optimal policy for each…
• Simulation
• Graphs

A preponderance of preprint servers

Having sent in my review for IEE, do I also post it on the peerJ preprint? File each complaint or suggestion as issue on the paper’s Github repository? Post a comment on the corresponding author’s blogpost on the paper? In my own online notebook? Oh, the reviewer’s plight! From review-recognition poverty to an embarassment of riches!

Other writing and thoughts

• Open Science cover letters?
• In which I rant about scientific software again

ews-review

• re-submitted. submmitted_revisions tag

Reading

• Joppa et. al. (2013) utoh… I rant about software again

• Are unreasonably harsh reviews retarding the pace of coral reef science? I applaud the spirit of this piece. When I saw the article, Bruno et al. (2013) appear on PeerJ last week I thought to myself, “Ooh, looks like controversial stuff! Bravo for using a preprint server,” so it is quite delightful to read the back story and confirm my guess. Though I am a proponent of open peer review, I find the practice of posting anonymous peer reviews a bit ethically murky still.

Misc

• Trying to submit knitcitations to CRAN; R-devel has updated linewrapping requirements that brake roxygen2 roxygen/125, so have to correct .Rd files after the fact.

• My ans: Open data vs Linked data on opendata.stackexchange opendata SO launches public beta

• Why people don’t use WADL Basically we can accomplish structure much better by having well-defined media types. We can define the schema of the json or xml file we return, ¹ and then the application knows what to expect and what it can do with the data. This is far more useful than just having a programmatic version of the API calls in the native language (“client objects”). Seems like REST was supposed to make such things obsolete, and to make deploying the API far simpler than it was in SOAP (which provides WSDL).

• An interesting discussion on using mediatypes for REST APIs, which might be referred to as the HATEOAS approach:

which basically means that responses from the server will be documents that include URIs to everything you can do next

hmm, a powerful idea.

References

• L. N. Joppa, G. McInerny, R. Harper, L. Salido, K. Takeda, K. O’Hara, D. Gavaghan, S. Emmott, (2013) Troubling Trends in Scientific Software Use. Science 340 10.1126/science.1231535
• Bruno et al. (2013) https://peerj.com/preprints/19/