FISH 6002: Week 12 - Interactive plots

Transcript

1. Week 12: Interactive plots FISH 6000: Science Communication for Fisheries Brett Favaro 2017 This work is licensed under a Creative Commons Attribution 4.0 International License
2. This week • Interactive plots with Shiny • Why? • How?
3. Why? - Explore complex data - Promote scientific curiosity and engagement - Rapidly test hypotheses - Promote experimentation, particularly among general public - Allow for real-time exploration of model output
4. Examples http://carboncounter.com/ https://github.com/d3/d3/wiki/Gallery https://rviews.rstudio.com/2017 /05/03/shiny-in-medicine/ http://www.tylernwolf.com/ ?utm_source=d3_org&utm_ medium=demo_link&utm_c ampaign=d3_website#/portf olio/corrdisp http://riskcalc.org/ ColorectalCancer/ https://canada.energypolicy.solutions/ https://getdolphins.com/blog/interactive-data-visualizations- new-york-times/ https://trase.earth/
5. https://shiny.rstudio.com/gallery/ Our tool:
7. Minimal example: Single File File → New File → Shiny Web App
10. ui = User Interface: All the stuff that collects input from the user Input types Layout Additional element in layout Output type
11. http://shiny.rstudio.com /images/shiny- cheatsheet.pdf
15. There is always a connection between input and output
16. renderPlot creates the plot space The rest of the code in the {} draws the plot itself Final code to execute the plot
17. The faithful vector is built-in to R This half sets up layout and collects user input… …while this half executes code and produces output (here, a plot)
18. Example 2:
19. To read in your own data: Embed the read.csv within your server code, and before renderPlot Note the use of relative file paths. Organize Shiny apps using R Projects
20. Example 2: All R stuff (e.g. loading data, packages) is on the SERVER side
21. Load package Load data Create a subset of the data, including only data with X values less than the value grabbed from the user interface (expressed as input$plotrange) Model X by Y from our subset Create a scatterplot with Y as a function of X Draw a straight line based on the linear model (i.e. a “line of best fit”)
23. Time = 15 min • Open the Minimal Example R Project • Change layout, input type, and output. E.g. • Layout: Make it a flow layout, split layout, or vertical layout • Input: Instead of a slider, make it multiple choice, • Output: Advanced - Try another output style Let’s modify the above code
24. Minimal example: Multiple File File → New File → Shiny Web App Divides code into two files ui.R server.R
25. library(shiny) # Define UI for application that draws a scatterplot ui <- fluidPage( # Application title titlePanel("Minimal Example"), # Sidebar with a slider input for x range sidebarLayout( sidebarPanel( sliderInput("plotrange", "Maximum X value:", min = 1, max = 100, value = 50) ), # Create a plot for the XY data mainPanel( plotOutput("xyPlot") ) ) ) # Define server logic required to draw a scatterplot with a simple linear model server <- function(input, output) { library(tidyverse) MinimalExample <- read.csv("../Data/MinimalExample.csv") output$xyPlot <- renderPlot({ datasubset <- MinimalExample %>% filter(X <= input$plotrange) model <- lm(Y~X, data=datasubset) plot(Y~X, data=datasubset) abline(model) }) } # Run the application shinyApp(ui = ui, server = server) app.R ui.R server.R library(shiny) ui <- fluidPage( # Application title titlePanel("Minimal Example"), # Sidebar with a slider input for x range sidebarLayout( sidebarPanel( sliderInput("plotrange", "Maximum X value:", min = 1, max = 100, value = 50) ), # Create a plot for the XY data mainPanel( plotOutput("xyPlot") ) ) ) library(shiny) server <- function(input, output) { library(tidyverse) MinimalExample <- read.csv("../Data/MinimalExample.csv") output$xyPlot <- renderPlot({ datasubset <- MinimalExample %>% filter(X <= input$plotrange) model <- lm(Y~X, data=datasubset) plot(Y~X, data=datasubset) abline(model) }) } One-file: Two-file: Both methods are technically equivalent. Do whatever works for you
26. Integrating Shiny into Markdown
28. Equivalent to ui.R Equivalent to server.R
30. i.e. both one-file and two-file are accepted
31. Real-world example: the cod rebuilder “Cod are coming back” “Cod will recover in 3-5 years” “DFO is lying about how many cod there are” How can we visualize the recovery of cod, so we can have a more informed public debate about the issue?
32. Goals: • Produce a simple visualization of the recovery of cod, given exponential growth at a rate specified by the user: • “I think cod will recover at X% per year”  show what this looks like (a curve) • Calculate how many years it will take for cod to grow beyond ‘critical’ abundance zone, given the user’s assumed growth rate • Allow the user to ‘correct’ for DFO’s “errors” – i.e. if you think there are TWICE as many fish in the sea as DFO says, let them put that into the model • Allow the user to set a custom recovery target (how many cod do we WANT there to be?) • Show this recovery in historical context (how many cod will there be vs, 100 years ago?) • Be transparent with all assumptions (i.e. extensive documentation)
43. These sliders populate four values: GrowthRate Target CorrectionFactor StartYear
44. input$ tells R to grab those values from the user interface Cod rebuilder code is long – let’s just focus on stuff pertaining to interactivity
46. Load packages and functions needed for second plot polyCurve function courtesy of: https://www.fromthebottomoftheheap.net/2013/01/11/shading-regions-under-a-curve/
48. Adjustments to any slider causes both plots to re-draw The growth function, polygon, height of red line, and X and Y ranges are all re-calculated for each plot
49. Use of RMarkdown allows for extensive documentation The Cod Rebuilder makes tons of assumptions – but it’s transparent about them
50. Sharing Shiny apps:
51. https://pandalusplatyceros.shinyapps.io/cod_rebuilder/ See https://twitter.com/S hinyappsRecent for recent Shiny apps
52. Interactive Plots recap • If you are R-literate, you are also interactive plots-literate • Any model, plot, etc. can be an interactive plot. If you are inputting data, you can get the user to input data instead – and then you have an interactive plot • This is incredibly powerful. • E.g. Cod rebuilder – don’t take MY word for it, play with the stock YOURSELF. What conditions do you have to meet to rebuild cod in 3-4 years. Do YOU think they’re realistic? Interactive plots allow for such experimentation • E.g. Stock assessment. I used an exponential growth model. There’s zero reason you couldn’t use a proper stock assessment model in this way. • Consider how your own data could be expressed as interactive plots
53. Course recap •You now have all the basic skills needed to work with data in R: •To manipulate it • To plot it • To share it •You can now be reproducible and transparent in your research!
54. Course recap •We discussed • Collecting data in the field, and cleaning data from a database • Data Policy – including data management plans • How to make figures of all kinds, ranging from the most basic to fairly complex, and making them publication- quality • How to save your data and scripts for future reference •You are now ready to do stats in R!
55. Thank you • Please consider doing a course review. All feedback welcome!