10/23/2017
The goal of the project is to let users explore interactively the Air Quality dataset for 2017 provided by EPA (source data here).
I've set up a Shiny application (direct link here) with three important components:
Leaflet to display geographical locations of air quality sensorsWe can download and process the air quality dataset, like below.
library(dplyr); library(lubridate)
download.file("https://aqs.epa.gov/aqsweb/airdata/annual_conc_by_monitor_2017.zip",
destfile = "annual_conc_by_monitor_2017.zip")
unzip("annual_conc_by_monitor_2017.zip")
anc_sub <- tbl_df(read.csv("annual_conc_by_monitor_2017.csv")) %>%
filter(Parameter.Name %in% c("PM2.5 - Local Conditions", "Ozone", "Sulfur dioxide")) %>%
mutate(X1st.Max.DateTime = ymd_hm(X1st.Max.DateTime))
And then apply calculations, like linear model, for a given subset.
my_model <- with(anc_sub, lm(X1st.Max.Value ~ X1st.Max.DateTime)) summary(my_model)$coefficients
## Estimate Std. Error t value Pr(>|t|) ## (Intercept) 4.412931e+03 6.801735e+02 6.487949 9.202738e-11 ## X1st.Max.DateTime -2.952874e-06 4.573397e-07 -6.456630 1.131112e-10
This is resource-consuming (run on our side) and code-heavy.
The user can easily interact with data, calculations are run on the server side.
Example: high sulfur dioxide values have been recorded near a hawaiian volcano.
renderUI() is a very useful function for generating input structures dynamically.# ui.R
selectInput(inputId = "sensor_type",
label = "A. Choose a sensor type:",
choices = c("PM2.5 - Local Conditions", "Ozone", "Sulfur dioxide")),
uiOutput("choose_measure")
#server.R
output$choose_measure <- renderUI({
range <- anc_sub %>% filter(Parameter.Name == input$sensor_type) %>% pull(Metric.Used)
selectInput("selected_measure", "Choose a measure:", unique(range))
})
Leaflet is available for Shiny and well integrated.