Do a quick analysis of the Health data collected by iPhones

health.R

library(XML)

# pull in the xml file
# can export this from the iPhone and e-mail it to youself
xx <- xmlParse("export.xml")
xx <- xmlToList(xx)

# now do some data mudging
library(plyr)

# grab only step counts
gg <- ldply(xx, function(x){
  if(!("type" %in% names(x))){
    return(NULL)
  }
  if(x[["type"]]=="HKQuantityTypeIdentifierStepCount"){
    return(x)
  }else{
    return(NULL)
  }
})

# remove some useless columns
gg$device <- gg$.id <- gg$type <- gg$sourceName <- gg$sourceVersion <-
    gg$unit <- NULL

# make the dates make sense
# not sure if this is right because it looks like everything is
# recorded in UTC, but I definitely wasn't in UTC all year...
library(lubridate)
gg$ymd <- ymd_hms(gg$startDate)

# just get 2017 data
gg <- gg[year(gg$ymd)==2017, ]

# truncate to days
gg$day <- as.character(trunc(gg$ymd, "days"))
gg$value <- as.numeric(gg$value)

# get steps/day
dd <- ddply(gg, .(day), summarize, steps=sum(value))

# shortcuts
dd$day <- ymd(dd$day)
dd$wday <- wday(dd$day, week_start=1, abbr=FALSE, label=TRUE)
dd$month <- month(dd$day, abbr=FALSE, label=TRUE)


# now do some plotting
library(ggplot2)

g <- ggplot(dd) +
  geom_col(aes(x=day, y=steps)) +
  geom_hline(yintercept=10000, colour="red") +
  theme_minimal() +
  labs(x="Date", y="Steps")

g_day <- ggplot(dd) +
  geom_boxplot(aes(x=wday, group=wday, y=steps)) +
  geom_hline(yintercept=10000, colour="red") +
  theme_minimal() +
  labs(x="Week day", y="Steps")


g_month <- ggplot(dd) +
  geom_boxplot(aes(x=month, group=month, y=steps)) +
  geom_hline(yintercept=10000, colour="red") +
  theme_minimal() +
  labs(x="Week day", y="Steps")