In an effort to improve my understanding of statistics, machine learning, and R, I undertook two projects this month: visualizing the danger of multiple t-test comparisons and calculating linear regression coefficients via gradient descent. With both projects, the emphasis was on ease of use and understanding: I created R functions that make it simple to manipulate parameters and visualize the results. All code is included in the respective GitHub repositories.
This month, I had the opportunity to speak with
I was one of the teaching assistants for a Princeton tropical ecology course in Panama. It was lots of fun snorkeling!
More thesis work. Still processing data for the live predation project, but also making some interesting progress in the analysis for the sensory networks project. Discussions continue with Pawel Romanczuk, Colin Twomey, and Bryan Daniels about a model for complex contagion.
I visited my advisor in Germany. The trip was both productive and fun! It’s great to see how the lab has developed. I presented on my live predation work to the group and had numerous discussions on another project involving quantifying sensory networks in fish schools. I also attended Damien Farine‘s social networks analysis workshop, which was informative. I’d expected the workshop to predominantly Max Planck students and was pleasantly surprised at how widespread the attendees’ home institutions were.
Thesis work continues as always. I volunteered on a short project with Statistics Without Borders to translate a WHO client’s workflow from Excel to R. The process was one of the first times I’ve programmed as part of a team, and the entire experience was immensely rewarding.
I reviewed a paper for Frontiers in Applied Mathematics and Statistics and wrote a blog post quantifying my daily commute. I spoke about my live predation work at one of the weekly seminars for the Integrated Behavioral Research Group at Princeton.
Work on the thesis chapters continues. I reviewed a paper for Behavioral Ecology and attended a few R Meetups, which was fun. I also wrote a few blog posts this month:
– How to quickly compare values between scalars, vectors, and matrices in R
– How to calculate rotation versus polarization (Couzin et al. 2002)
– Revisiting the computational efficiency of for loops versus apply
This month saw a lot of progress on the Schreckstoff perceived risk project: I’m collaborating with Pawel Romanczuk and colleagues to combine data and theory on how fish schools perceive and encode risk in their environment. There’s a lot of details to sort through and decide how to distribute among papers, and plenty of writing to fine-tune before we can send it out, but the results are looking interesting and I’m excited to share them within the next few months. Data processing continued for two other projects (videos of the trials are just the first step!), and I wrote a reflections piece on the fourth year of my PhD.
August saw the PhD defenses of two Couzin labbers: Mircea Davidescu and Andrew Hartnett, pictured below. (Princeton EEB also watched Cara Brook, Charlotte Chang, Josh Daskin, Lisa McManus, and Andrew Tilman defend. Busy month!) I reviewed a paper for New Journal of Physics, which was useful practice for thinking more quantitatively, and I conducted the second round of predator-prey experiments. I also went on Princeton’s student-run radio station WPRB to talk about the process of science with good friend Stevie Bergman.
I started preparing for round two of predator-prey experiments, which involved picking up and acclimating more fish. I’ve found every visit to the hatcheries fascinating: I spend so much time staring at shiners that it never ceases to make me laugh and watch in awe when I see fish that are vastly different in shape or size, and to have thousands of them hanging out together in a massive tank. The transition from hatchery to lab life went well, largely thanks to years of fine-tuning the process for the first hours and days the fish are in lab. I also gave lab meeting with most of the group in Konstanz and me in New Jersey, which involved sharing my screen on Google Hangouts. It was a bit of a strange experience to give a talk through my computer, but the feedback was very useful. I’m hoping to finish up analyses on the Schreckstoff project soon and I’m almost done with a first draft of the manuscript. I’m looking forward to sharing the results of years of work – hopefully soon.
I’m continuing to run experiments, but I found time to enjoy Princeton Reunions, which were a blast as always. I was saddened to hear that NSF discontinued the DDIG program for Integrative Organismal Systems and Environmental Biology. The experiments I’m currently running are funded by a DDIG, and I greatly value the independence this grant has given me and the practice in managing a budget and applying for funding. Otherwise, I’ve enjoyed having two excellent undergrads working for me over the past few weeks, and I’ve been tutoring them in statistics and modeling. We’re now recreating my advisor Iain Couzin’s 2002 agent-based zonal model for fish schooling in R.
Life kicked into a higher gear as I began experiments I’ve been thinking about for years now. Looking forward to sharing exciting results at some point, but for now here’s a snapshot from one of the trials.
Outside of experiments, I started the Science Life tab on this website and rewrote the first Introduction to R blog post to utilize WordPress’s fancy syntax highlighting. Princeton hosted students from Humboldt University in a workshop on social modulation of risk (from empirical biology to theory, psychology, and economics!) and it was a major success.
It was a busy month. As a member of IACUC, I helped conduct semiannual inspections of all laboratory facilities at Princeton. Simon Levin gave a memorable talk about the history and future directions of mathematical models in ecology. Jacob Davidson, a postdoc with Iain in Konstanz, visited us in NJ and we had great discussions. Thanks to DDIG funding, I hired a team of undergrads to help with manual correction of fish trajectory data and preparation for summer experiments. They have been a huge help. I started lessons on statistics and R with them, which has been a great review for me, too. In preparation for upcoming experiments, the lab received two batches of hundreds of shiners each and Colin Twomey, Joe Bak-Coleman, and I were busy keeping them healthy and happy as they transitioned from Arkansas hatchery life to the lab. I delivered my fourth-year talk to the EEB department, which went well and gave me good feedback. Here are a few slides from the talk:
Not much to report for this quiet month. I’m gearing up for predation experiments over the summer, which involves coordinating with a NJ fish hatchery, Laboratory Animal Resources at Princeton, a tank supplier, EEB administration, and potential undergrads I’m looking to hire. I submitted a second round of review for an article, and am reading the literature and continuing analyses in R. I’ve been reading A Biologist’s Guide to Mathematical Modeling in Ecology and Evolution (Otto & Day 2007), a great resource.
I received an NSF-DDIG (National Science Foundation Doctoral Dissertation Improvement Grant) this month, something I’m excited about and very grateful for. I keep chugging along with analyses for the Schreckstoff project. I’m looking forward to sharing results at my fourth-year talk at Princeton in April.
I’m at the point in the thesis where it’s time to pull out the textbooks and re-learn some stats. Coincidentally, re-learning the within- vs. between-group variance distinctions in ANOVA helped me discover an answer to a long-standing question I’ve had in categorizing fish startle behavior. The k-means clustering algorithm divides your data such that the within-cluster sum of squares is minimized and the between-cluster sum of squares is maximized. You can take it a step further with a Gaussian mixture model, which allows for different within-cluster sums of squares. The mixture model worked beautifully on my data. I’ll continue tinkering with the analysis to ensure the categorization works as well as I’d like it to.
Princeton’s radio station, WPRB, welcomed me back with a great discussion with plasma physics PhD candidate Brian Kraus about public perceptions of science (note: link to audio might not be visible on a Firefox browser). We discussed public distrust of science, social media echo chambers, and the role scientists have in communicating their results beyond the ivory tower. Coincidentally, Nature published an article the next day on how the only way scientists can “change the world” is by engaging non-scientists. For the thesis, I finished processing videos of fish startle events. Now for analysis!
This month was broken up with a wedding, the Election, and Thanksgiving. Donald Trump’s victory motivated me to write a piece titled “A call for self-evaluation, critical thinking, and empathy” where I write about the massive political divide in America, synthesize liberal confusion and frustration, and think about how Americans (myself included) can move forward under the new administration. One big step, I think, is understanding white working class voters, a group that Trump galvanized. I’ve begun reading Arlie Russell Horschild’s Strangers in their Own Land to gain a better understanding of the Tea Party as well. As for the thesis, I’m plowing away at video processing and have begun analyzing the data as they come in. The data consist of startle cascades, or socially contagious bursts of fast swimming in fish schools.
October began with the PhD defenses of Couzin lab members Colin Twomey and Ari Strandburg-Peshkin, two fantastic scientists and big inspirations for me. I submitted the DDIG, had good discussions with former Couzin labber Pawel Romanczuk (now at Humboldt University, Berlin) and the mechanical and aerospace engineers in Naomi Leonard‘s group here at Princeton, and I reviewed two articles. With Open Labs, I had fun talking science with 10-year olds at the Young Scholars’ Institute in Trenton and around 100 local high schoolers at our third on-campus Science Cafe. Thanks to Aida Behmard for co-organizing and to the other Open Labs speakers for making it possible!
This month was much of the same as before: experimental pilots and data processing. I began writing an application for the NSF-DDIG, a grant for PhD candidates who have passed their generals exam. Creating figures for the application involved plenty of R, which I really love. A few posts for The Headbanging Behaviorist R blog series have been in the back of my mind for months now and I’m hoping to find the time soon to crank them out. September also involved coordinating a few upcoming Open Labs events at Princeton and Trenton.
August was productive! A good chunk of the month was spent processing videos for a set of experiments I conducted in May. I had a lot of fun with physics PhD candidate Stevie Bergman on her show These Vibes are Too Cosmic – a show on Princeton’s WPRB radio station – where she interviewed me about my research and collective behavior in general. Finally, at the end of the month and culminating in a year of work, I began pilots for a predator-prey experiment in the lab. It’s all been very exciting, and I’m glad to have the help of Geoffrey Mazué, who was instrumental in the fish work back in January. It’s been great to have Joe Bak-Coleman and Jolle Jolles on the project, as well, to make sure we get the most interesting and ecologically-relevant results possible from the experiments. Finally, I wrote a reflections piece on the 3rd year of my PhD, a time when a lot of my perspective changed.
Most of July was spent with the Couzin Lab in Germany. I attended the ECBB conference in Vienna, which was fun but reminded me of how important quantitative rigor is. I got the chance to speak with Lucy Aplin, one of my favorite researchers, and to meet Wytham Woods mastermind Ben Sheldon as well. I gained IACUC approval to conduct a set of experiments I’m excited about, and when I returned to the States, I worked on all the loose ends that comes with starting a new study system. For Highwire Earth, I wrote a post on the UN’s second sustainable development goal: ending world hunger.
June has been a mix of vacation, video processing, and great discussion. After experiments concluded last month, I began the long process of tracking fish in the videos and ensuring the tracks are perfect. For videos longer than a few seconds, and for schools with 40 fish, this can take a while! Once the tracks are complete and run through visual field reconstruction software (courtesy of the brilliant Colin Twomey), the really exciting step begins: analysis. This is when I love being a scientist; the fish begin speaking back to you through the data, and you get to share excitement and ideas with other scientists. I spent the second half of June in Konstanz, Germany, with the German half of the Couzin lab. Before heading across the ocean, I attended the Kappa Alpha Psi Fraternity robot car race, where Trenton middle and high school students raced cars they had spent weeks building. It was a lot of fun! I also had the privilege of seeing the current generation of Kappa Alpha Psi high school seniors announce their college choices and sign oaths of excellence.
Data collection for a good chunk of the thesis is finally complete! I took a much-needed vacation after working essentially every day of 2016 on experiments. (Fish don’t take weekends off…) Now video processing and tracking begins, which should take a few weeks. May was a big month for Open Labs, where we met with the heads of the Young Scholars Institute and the I am Trenton foundation to begin a series of talks with Trenton students in after-school teen programs. Several members then gave talks on two occasions to the Trenton Housing Authority robotics club, and we’re continuing to meet with students over the summer. Open Labs also hosted its second Science Cafe for Princeton-area high school students, which featured talks on modeling, civil engineering and architecture, collective behavior, and exoplanets. Finally, the blog posts on R continued with an entry on writing functions and if statements.
This was a busy month! I continued to collect and analyze data for thesis experiments on alarm propagation in fish schools; I’ve begun using visual field software developed by colleague Colin Twomey to get at the sensory network structure of the group. We also began collaborating on a project involving fission-fusion dynamics under different environmental conditions. I gave a talk at the annual Princeton-Penn-Rutgers-Columbia graduate student conference, which was in Philly this year and was a lot of fun. A few weeks later, Julio Herrera Estrada and I spoke about Highwire Earth on WPRB’s “These Vibes are too Cosmic” radio show. It was a live broadcast so I was definitely nervous, but it went well and we had a blast. Open Labs helped judge the yearly science fair for 3rd through 5th graders at Hopewell Elementary School, and we’re looking forward to upcoming events. Finally, I at last got around to continuing the blog posts on R: I published one last week on for loops and random walks.
Starting one experiment on collective antipredator behavior in fish and setting up for the second half of the weeks-long experiment. I helped organize and gave a few talks at the first Princeton Open Labs Science Cafe, where we invited local high school students to hear talks on grad student research and what it’s like to be a scientist. I also started working with an enthusiastic undergrad and an ambitious high school student. The spring flowers have started to bloom, making the morning commute a lot prettier!
Wrapped up the first half of a weeks-long experiment. Habituation can take forever. A preliminary look at the videos shows there might be something interesting here, but quantifying it and getting at the mechanism looks challenging. Meanwhile, I helped organize a statistics workshop series for the Ecology and Evolutionary Biology grad students, which is a much-needed sharing of statistical knowledge for all of us. I also helped launch Highwire Earth and organized and helped run the first Princeton Open Labs event at Hopewell Elementary School! The 5th graders are so much smarter and curious than I could have anticipated.
I got a huge surprise when we opened an order for 300 fish and found… 1200 fish instead. With some help and creativity from two Couzin labbers, we managed to give all the fish a home and keep them healthy and happy for experiments. In between a hectic schedule of experimental trials, I also taught a short course on science communication for Princeton undergrads and grad students. It was a lot of fun!
Cover photo credit: Eduardo Lopez Negrete