Lecture 18: Interactive plots with ggvis

STAT598z: Intro. to computing for statistics

---

Vinayak Rao

Department of Statistics, Purdue University

ggvis is a simple way to get interactive plots

• provides a simpler interface to shiny
• is still experimental

Like ggplot this expects a dataframe/tibble as an input

Some differences:

• add layes using %>% instead of +
• instead of aes(color=group), write color = ~group
• we still write color:=clr_val
• aesthetics have different names:
  • color becomes stroke
  • alpha becomes opacity

library('tidyverse')
library('shiny')
library('ggvis')
load('HomeValues.RData')
HomeValues$qtr <- as.double(HomeValues$qtr)

Attaching package: ‘ggvis’

The following object is masked from ‘package:ggplot2’:

    resolution

plt<- ggvis(HomeValues,x=~qtr,y=~Home.Value,stroke=~State) %>% 
        layer_lines(); plt

plt <- plt %>% hide_legend('stroke'); plt

plt %>%  layer_points(); plt

plt %>%  layer_points(size=1, fillOpacity=.1) # Bug!

ggvis uses both = and := for assignments

Use = to map a variable to a property

• Then use ~ to refer to a column of a dataframe

Use := when we set a property based on a value

plt %>%  layer_points(size:=1, fillOpacity:=.1)

In the end, set properties using = ~column or := value

So why use ggvis instead of ggplot?

• Interactive plots!

plt %>% layer_points(size:=input_slider(.1,50,1), 
                     fill =~ State, fillOpacity:=.5) %>% 
         hide_legend('fill')

plt  %>%  add_tooltip(function(x) 
           {paste(x$State,":",x$Home.Value)},'hover')

add_tooltip needs a function to read value and return a string

• we used an anonymous function to print State,Value

For lines, add_tooltip only prints first value (http://stackoverflow.com/questions/28540504/mouse-hover-in-layer-lines-ggvis-r)

• add layer_points() for all values

plt <- plt  %>% layer_points(size:=input_slider(0,5))

plt  %>%  add_tooltip(function(x) 
                {paste(x$State,":",x$Home.Value)},'hover')

mtcars %>% ggvis(~wt, ~mpg) %>% layer_points() %>%
  layer_smooths(span = input_slider(0.2, 1))

plt <-ggvis(HomeValues x=~qtr,y=~Home.Value,stroke=~State) %>% 
        layer_smooths(span:=input_slider(0,5)) %>% 
        hide_legend('stroke')

Error in parse(text = x, srcfile = src): <text>:1:24: unexpected symbol
1: plt <-ggvis(HomeValues x
                           ^
Traceback:

Error because ggvis doesn't do grouping for you (unlike ggplot)

plt <- HomeValues %>% group_by(State) %>% 
       ggvis(x=~qtr,y=~Home.Value, stroke=~State) %>% 
        layer_smooths(span=input_slider(0,2,step=.1)) %>% 
        hide_legend('stroke')

ggvis(HomeValues) %>% layer_histograms(x=~Home.Value)

ggvis(HomeValues) %>% 
  layer_histograms(x=~Home.Value, 
                   width=input_slider(min=1000,max=100000))

ggvis(HomeValues) %>% 
  layer_histograms(x=~Home.Value, fill.hover:='red',
                   width=input_slider(min=10^3,max=10^5))

plt<- ggvis(HomeValues,x=~qtr,y=~Home.Value,stroke=~State) %>% 
        layer_lines(stroke.hover:='black') %>% 
        hide_legend('stroke')
plt %>%  add_tooltip(function(x) 
             {paste(x$State,":",x$Home.Value)},'hover')

tidyverse commands can be overloaded for use with ggvis:

https://rdrr.io/cran/ggvis/man/dplyr-ggvis.html

plt <- HomeValues %>% group_by(State) %>%  
        ggvis(x=~qtr,y=~Home.Value, stroke=~State) %>% 
       filter(State %in% eval(input_select(choices = 
                        unique(as.character(HomeValues$State)), 
                 multiple=TRUE, label='States list'))) %>% 
                   layer_lines(strokeWidth:=2)

Note the eval, this is because of we are calling input_select inside filter

http://stackoverflow.com/questions/25891020/dynamic-filtering-with-input-select-using-ggvis-in-r

library('ggplot2');library('maps') 
my_state_map <- map_data('state');
my_state_map$region <- tolower(my_state_map$region)
get_ab <- function(x) state.abb[x == tolower(state.name)]

get_house_pr <- function(st,yr) {
    HomeValues[HomeValues$State==st &  HomeValues$qtr==yr,2] }

state.name[51]<-"district of columbia"; state.abb[51]<-"DC"
# apply get_ab to each row of my_state_map
my_state_map$region <- purrr::map_chr(my_state_map$region, 
                                      get_ab) 
get_yr_pr <- function(yr) { # Function to get vector of prices
  pr <- my_state_map$pr     #  of yr
  for(st in state.abb) 
    pr[my_state_map$region == st] <- get_house_pr(st,floor(yr))
  return(pr) 
}

Attaching package: ‘maps’

The following object is masked from ‘package:purrr’:

    map

my_state_map  %>% ggvis(~long, ~lat) %>%  
   mutate(pr = eval(input_slider(1976,2013, 
         map= function(x) get_yr_pr(x)))) %>% 
   group_by(region) %>% layer_paths(fill=~pr)  %>%
   hide_axis("x") %>% hide_axis("y") %>%
   set_options(duration=0,width=960,height=600,keep_aspect=TRUE) %>% 
   hide_legend('fill') %>%  hide_legend('stroke') %>%  
   add_tooltip(function(x) {
       paste(isolate(x$curr),":",x$region,":",x$pr)},'hover') %>%
   scale_numeric("fill", range = c("yellow","red"))

ggvis is also compatible with reactive programing

This is a programming paradigm imported from shiny (https://shiny.rstudio.com/articles/reactivity-overview.html)

At a high level a reactive source feeds inputs to reactive end-points

• whenever the source changes, the end point is automatically updated

ggvis automatically updates when a reactive input changes

#https://r2014-mtp.sciencesconf.org/file/92631
#library(shiny)
dat <- data.frame(time=1:10, value=runif(10))

# Create a reactive that returns a data frame, adding a new
# row every 2 seconds
ddat <- reactive({
  invalidateLater(500, NULL) # wait of 2 seconds
  dat$time <<- c(dat$time[-1], dat$time[length(dat$time)] + 1)
  dat$value <<- c(dat$value[-1], runif(1))
  dat
})

ddat %>% ggvis(x = ~time, y = ~value, key := ~time) %>%
  layer_points() %>% layer_paths()