Wednesday

R packages

  • Trying to build Windows binaries for R packages using win-builder service.

  • First have to get R CMD check working.  Check runs in –vanilla mode so have to tell it where libraries are.  Created the file ~/.R/check.Renviron containing the line: R_LIBS=“/home/cboettig/R/packages” seems to do the trick.

  • Successfully built windows binaries for RMendeley package (pre-release, still in development). Download the binaries from github.

  • Matt successfully installed CUDA on zero (2.6, looks like we might want 3.0 or 4.0 – if only they could organize their site better…).  Now, do we pay for an academic license of CULA tools and start playing with gputools?  A few more specialized GPU R packages are on the HPC taskview.

Warning Signals

  • Troubleshooting MCMC – resolved some minor issues from last night’s run.  Seems to be up and running now (on zero, pid 14301, nice 5), we’ll see if we get successful output…

Re-running with smaller step increments on m (1/20th of value), see if the posterior is smaller). Running with four chains, still 1e5 steps, nice 18.
Running with four chains, 1e6 steps, nice 19.

  • Running longer versions of sampling effect (have enough replicates to get smooth distributions), on zero, nice 19

  • Can represent these as ROC Curves, showing the result of 20, 50, 100, and 200 data points.

For deteriorating example in this dataset

  • Working on manuscript edits, 4:30p - ??

Teaching

Teaching second half of phylogenetics comparative methods tutorial in R in core today.  Using my Bodega curriculum, covered up through model choice.  3p-4:30p.

Reading

  • Ooh: Stack et. al. 2011

  • Losos’s address from Evolution last year, out in Am Nat.   (Losos, 2011)

  • Cass pulled out this gem from Bruce Alberts: (Alberts, 2005)

  • Daniel says peer review should be public (Mietchen, 2011)

  • Wow, once again Jeremy really writes about some great stuff on Oikos blog. Just a read a piece after my own heart on the meaning of “random” in ecology. I think I need to cut back, I might be addicted.

To Do

Things I haven’t gotten to today. Should add these to their github repositories issues trackers.

  • Dryad R package – search metadata. Scott is taking a whack at this one.

  • Treebase support for downloading character matrices. Should be quick. one day

  • Treebase writeup

  • SocialR – integrate Rflickr, twitteR, begin github api, improved tracking of relevant code and data

References