Monday: some code-tricks, Algorithms group planning

Teary

Some discussion on combining Greek letters with variables in R.  Previously have solved this with substitute() command,

x = 4.5
plot(1:10, main = substitute(paste("Kendall ", tau == val), list(val = x[1])))

but Peter has a more elegant solution that can be vectorized:

sapply(interaction("sigma",1:5, sep=" == "),function(x)parse(text=x))
#for instance
plot(0)
legend("topright",legend=sapply(interaction("sigma",1:5, sep=" == "),function(x)parse(text=x)),lty=2)

Better yet, bquote has a mechanism to evaluate an argument vs express the argument as a symbol, see gist id=1108159.

  • Having other people to help code is great, we’re considering the teary list-serve as a potential source for R help now.
  • Also very interesting discussion around the NCEAS model, it’s impact, and the role of ecoinformatics. Check out UC Davis’s ICE lab.
  • Update comments/suggestions about Rblogging and CiTO, posted to Dave Munger. (not actually during teary)

Algorithms group

Our group is gaining interest among others, so we spend most of today thinking about our goals and possible changes to the format that could address this while preserving our monkey-style approach of low preparation, liberal doses of mistakes, and actually hacking out working examples of trivial problems together. We also started reviewing the set up for the EM problem, which we’ll consider when Yaniv returns the week after next. Next week we’ll detour into Amazon EC2 tutorial instead.

  • Nick has some great notes on the future of the session, which hopefully will be posted on the OWW page eventually. Meanwhile:

The Algorithm Discussion Group began as a handful of students who met informally to implement toy models of various algorithms in R. Every week, ADG members gather in the Schreiber lab to discuss an algorithm, create code to run a sample problem, and interpret the results. Over the course of the last two months, we’ve covered ABC and (MC)^3, and plan to cover EM and HMM in the near future. While we talk about reading papers beforehand and coming to meetings prepared, it often devolves into shouting instructions at the computer monkey.

Several students have expressed interest in joining our group over the last two months. Unfortunately, our current format would be suboptimal and unwieldy for groups with more than five or six members. We are instituting a few changes to open the group to new members and have three involvement options for interested students.

Sign up for the mailing list with Alisa (). We envision this as a way for students to talk about algorithms, mathematics, statistics, and coding. Most importantly, it’s going to be a safe environment to ask questions and not worry about sounding goofy.

Attend formal Algorithm Discussion Group meetings. Meetings will occur approximately monthly and be announced on the ADG mailing list. One student will lead the meetings and provide reading material and code prior to the event. The meetings will be interactive and accessible for students with beginning coding and rusty mathematical skills.

The third option is to pitch algorithm meeting ideas to the organizing committee. Interested students can suggest algorithms with general applicability to Pop Bio student research and have the option of implementing the algorithm in R with the organizing committee and presenting it at the monthly meeting. Currently, the organizing committee is Alisa, Carl, Dave, Nick, and Yaniv, but we envision members rotating in and out as commitments change and students leave Davis.

Anyway, reply to Alisa () if you’re interested in participating further. This will be the last email to the CPB-student list and all future emails will be to the ADG list.

  • updates to MCMC (with examples), and Dave pointed this intriguing example of bad behavior in MCMC convergence out.  Seems to be a rather different problem than I get from the priors, as it is rather contrived with a bimodal distribution to stump the particular centered prior, but oh well.

Warning Signals

Only a bit of real research today.  Updates to warningsignals, see git log.

Discussion for GTC

Great discussion about better alternatives to the cookbook lab report: 1. Moskovitz, C. & Kellogg, D. Science education. Inquiry-based writing in the laboratory course. Science (New York, N.Y.) 332, 919-20(2011).

  1. Asai, D.J.Measuring student development. Science (New York, N.Y.) 332, 895(2011).  Particularly interesting in that a big player is taking accurate assessment the effectiveness of educational impact seriously; something other funders (i.e. NSF broader impacts) have not done, and also in its reference to this excellent study characterizing gender and race/ethnicity differences in staying in the sciences.

  2. Deslauriers, L., Schelew, E. & Wieman, C.Improved Learning in a Large-Enrollment Physics Class. Science 332, 862-864(2011), also discussed in this NY Times piece.

Also looked into some work by Scott Freeman. Scott also sent the preprint for #2, which isn’t out yet, hence why I couldn’t find it.

  1. Freeman, S. & Parks, J.W. How accurate is peer grading? CBE life sciences education 9, 482-8(2010).
  2. Freeman, S., Haak, D. & Wenderoth, M.P. Increased Course Structure Improves Performance in Introductory Biology. Education (2011).doi:10.1187/cbe.10-08-0105
  3. Entry and Persistence of Women and Minorities in College Science and Engineering Education Entry and Persistence of Women and Minorities in College Science and Engineering Education. Educational Research (2000).
  4. Freeman, S. et al. Prescribed Active Learning Increases Performance in Introductory Biology. Education 6, 132-139(2007).