2021 Pre-Recorded Talks

Digital tracks will be released June 4

Andrew Li

University of British Columbia: Vancouver, BC

shinyRGT: An R-Shiny application for extraction and analysis of rat gambling task data

The Rat Gambling Task (RGT) is a well validated rodent model of addiction-like behaviour. It is based on the Iowa Gambling Task (IGT) - a commonly used clinical assay to measure gambling-like behaviour. Rats choose between 4 options to earn as many sugar pellets as possible within 30 min. Each option is associated with different reward sizes, but also the probability and duration of the time out punishment. The task is designed such that the optimal strategy for earning sugar pellets is to favour the low risk, low reward options. Consistently selecting the high risk, high reward options results in longer and more frequent time-out penalties. Currently, there is no specialized graphical user interface (GUI) designed to extract, clean, and process RGT data. The installation and use of existing tools are challenging for users lacking coding experience and can be extremely time consuming. To address these issues, we developed a free and open source R-Shiny application called shinyRGT, as a GUI for RGT data extraction, analysis, and visualization. Clean and usable data can be easily extracted. As well, publication ready plots can be readily generated and annotated from user input. All generated tables can be downloaded as CSV files and generated graphs can be saved to local machines. shinyRGT is deployed at https://andrewcli.shinyapps.io/shinyRGT/ for online use. The repository is available at https://github.com/andr3wli/shinyRGT.

Bio: Andrew is an incoming psychology masters student at the University of British Columbia under the supervision of Prof. Jiaying Zhao. He is interested in information visualization, human-computer interaction, and cognitive science. As well, he develops and teaches R workshops to undergraduate students. In his free time, Andrew enjoys keeping up with the latest tech, golfing, and training for his first marathon.

Diana Dishman

Pronouns: she/her
NOAA Fisheries West Coast Region, Protected Resources Division, Portland Branch

Helping Regulatory Teams Work Better, Together

NOAA Fisheries is responsible for administering the Endangered Species Act (ESA) to insure actions by any government agency are not likely to jeopardize the continued existence of any threatened or endangered species we manage. In the West Coast Region this includes many species of whales, sea turtles, sea lions, salmon, steelhead trout, and marine fishes. Our region’s biologists implement ESA policy across Washington, Oregon, Idaho, and California and are responsible for analyzing how proposed projects may impact protected species. The regulated community we serve (i.e. federal agencies, and any organization that receives funding or requires a permit from any federal agency) counts on NOAA Fisheries to provide authorization under the ESA to be able to carry out their projects. Our agency is also obligated to make sure our analyses are transparent to the public and based on the best available science.

Over time, our offices have had to find ways to complete more of these analyses within regulatory deadlines with fewer staff, all while maintaining the consistency of our methods and decisions. Adding to this challenge is a legacy of organizational culture (common to many federal agencies) where analysts tend to work in separate ‘silos’ specific to species, geographies, or project types, and communication across offices is limited. This situation presents our teams with a common problem: how do we perform rigorous analyses consistently, transparently, and on time with growing demand and limited resources?

Our team of regulatory biologists took a new approach to what were previously separate state-level evaluations, and used R to automate analyses of the impacts of research projects across the West Coast Region. In describing our process, this example will show how the new approach has transformed the workflow we use to analyze our data and generate authorizing documents. We now tackle the analysis as a collaborative team of code builders, code users, and output consumers, and this strategy has saved hundreds of collective staff hours and improved the consistency of our results. The approach also allows our team to provide ESA coverage to applicants faster, providing researchers greater confidence their projects can start when planned. Seeing how using R has improved our group’s efficiency and ability to balance workloads has already inspired other teams in our region to start adopting similar approaches. Government resource managers and policy analysts may not think of themselves as prime candidates for using code, but we are quickly learning R can set regulatory teams on a path to more efficient and flexible collaboration.

Bio: Diana Dishman is a Natural Resource Management Specialist in the Protected Resources Division of NOAA Fisheries’ West Coast Region. She has a Master’s in Biology from Portland State University, where her research focused on marine mammal population genetics, and a Bachelor’s in Biology from Scripps College. Prior to joining NOAA Diana worked in an aquatic toxicology laboratory looking at the impacts of contaminated waters on fish and invertebrates. She later fell in love with data management, analysis, and visualization as an environmental consultant, investigating contaminant and biodiversity data associated with Superfund sites, the Deepwater Horizon oil spill, and litigated water disputes, and working to distill complex data into clear and compelling products used to guide clean-up and recovery of impacted habitats. In her current role with NOAA Fisheries Diana is helping her branch streamline Endangered Species Act consultations and permitting by building products in R, and working to expand internal R training for others in the Region. Diana lives in Clackamas, Oregon, and when she's not coding is usually trying to keep up with her two daughters, two dogs, and too many farm animals.

Kim Gaines-Munkvold

Pronouns: she/her
Portland State University, Portland, OR

Session: Digital

Supervised Land Cover Prediction Waldo Lake Wilderness Area

Land cover classification maps provide useful information related to monitoring and managing vegetation, agriculture, natural resources, and urban planning. The purpose of this prediction is to identify burned areas in the Waldo Lake Wilderness Area. On August 23, 1996, the Charlton Butte Fire was started by lightning and burned 10,400 acres of the Waldo Lake Wilderness. Fire recovery has been slow in the nearly 25 years since the fire. Two models, Ranger and K-Nearest Neighbor, from the Caret package, training points, and Landsat 8 imagery were used to create the landcover prediction. A total of 7,500 training points were used to define five classes, including burned area, forest, low vegetation, water, and road. The results of this initial modeling contained noise from the road classification. This classification was removed quickly and easily from the final models to better visualize the burn area and vegetation. Both the Ranger and K-Nearest Neighbor models performed well, with Kappa scores ranging between 91 - 94%. These models were then tested in an untrained study area in which a fire had occurred. This presentation is supported by slides and leaflet web maps.

Bio: Kim Gaines-Munkvold recently received a GIS Graduate Certificate from Portland State University. Her work focuses on remote sensing and fire recovery. As a native Oregonian, her favorite place to perform spatial analysis is her own backyard in Portland, or her favorite camping spot, Waldo Lake in Central Oregon. During her time at Portland State, Kim became fluent in a variety of geospatial processing programs and programming languages; of these, she has found that R is one of the most flexible, and thus most valuable, in her skill set. She continues to refine her skills as she partners with other PSU alumni on mapping projects.

Matthew Bayly

Vancouver, British Colombia

Scheduled Web-scraping With R from a Server

It is likely that many of you have experimented with rvest and related packages for accessing (or scraping) data directly from a website, but what about performing these operations as routine tasks on a fixed schedule? For example, say you wanted to download avalanche risk forecasts and weather data at 6:00 am daily from various websites and store this information for subsequent analyses. We can accomplish this by running a given R script as a cron job (Linux/Mac) or by using the Task Scheduler (Windows) from either a desktop or server environment. This simple design pattern opens a whole world of possibilities such as enabling us to build our own real-time Shiny-free dynamic html/js dashboards by overwriting a single data file. I will also briefly discuss web scraping ethics, API endpoints and the polite.

Bio: Matthew Bayly works as a Decision Support Tool developer with ESSA in Vancouver. His work focuses on integrating information from various models, databases and real-time sensors into dynamic tools to support data visualizations and decision making. His broad background in environmental science and passion for ecological research fuels his interest in programming and web development as applied tools. Matthew enjoys the creative freedom and strategic design elements of programming in R. He also holds big visions for how all of this will revolutionize environmental management and decision-making writ large. His key focal areas include stream networks, fisheries, and water management.

Peter Boyd

Pronouns: he/him
Oregon State University: Corvalis, OR

Modeling spatio-temporal data with the nphawkes package

This talk will introduce R users to the flexible deployment of nonparametric Hawkes, or self-exciting, point process models by offering a guide to our new nphawkes package. Hawkes processes quite literally encapsulate an “infectious idea” as they are used to describe data that may be thought to spread through a self-exciting or contagion mechanism. As the literature on Hawkes processes grows, the use of such models continues to expand, encompassing a wide array of applications such as earthquakes, disease spread, social networks, neuron activity, and mass shootings. As new implementations are explored, correctly parameterizing the model is difficult with a dearth of field-specific research on parameter values, thus creating the need for nonparametric models. The model independent stochastic declustering (MISD) algorithm accomplishes this task through a complex, computationally expensive algorithm. In the package nphawkes, we have employed Rcpp functionalities to create a quick and user-friendly approach to MISD. The nphawkes R package allows users to analyze data in time or space-time, with or without a mark covariate, such as the magnitude of an earthquake. We demonstrate the use of such models on an earthquake catalog and highlight some features of the package such as using stationary/nonstationary background rates, model fitting, visualizations, and model diagnostics.

Bio: I am a PhD candidate at Oregon State University with a research focus on nonparametric Hawkes process models with applications to earthquakes as well as mass shootings. My interests include environmental statistics, backpacking, surfing, and growing copious amounts of kale in my garden.