---
title: "Health stats"
output:
  html_document: 
    toc: yes
  pdf_document: default
  html_notebook: 
    code_folding: hide
    toc: yes
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = FALSE, warning = FALSE)
```

# Summary
This documents provides an overview of health data recorded by *Gadgetbridge* using a *Pebble* smartwatch.

# Preparation
Put the *GadgetBridge* database in the current folder, then knit the current `.Rmd` file.

You should also set you **timezone** in the corresponding variable.

  * `PEBBLE_HEALTH_ACTIVITY_SAMPLE` is a simple table containing steps and timestamps.
  * `PEBBLE_HEALTH_ACTIVITY_OVERLAY` is more complex.  It tracks activities (sleep, deep sleep, nap…) with a start and an end date.  Note that date may overlap: you can be both in sleep and deep sleep.
  
I have decided that any event happening after 8pm is registered for the next day.  For instance:

  * if you go to bed at 9pm on Tuesday and wake up at 7am on Wednesday, the data will return 10 hours of sleep on Wednesday
  * if you go to bed at 7pm on Tuesday, a part of your sleep will be added to Tuesday.  The cutoff won't exactly be 8pm, it will depend on the duration of the first sleep session recorded by you watch.

```{r include=FALSE}
# Load data ------------------------------------------------------------------
library(DBI)
library(RSQLite)
library(tidyverse)
library(lubridate)
library(scales)
con <- DBI::dbConnect(RSQLite::SQLite(), dbname = "gadgetbridge")
steps_data <- dbReadTable(con, "PEBBLE_HEALTH_ACTIVITY_SAMPLE")
sleep_data <- dbReadTable(con, "PEBBLE_HEALTH_ACTIVITY_OVERLAY")
dbDisconnect(con)
#str(steps_data)

# Transform data -------------------------------------------------------------

## Interprate timestamp
timezone = "Europe/Paris"
steps_data$datetime      <- as_datetime(steps_data$TIMESTAMP)
sleep_data$datetime_from <- with_tz(as_datetime(sleep_data$TIMESTAMP_FROM), timezone)
sleep_data$datetime_to   <- with_tz(as_datetime(sleep_data$TIMESTAMP_TO),   timezone)

## Calculate activity duration
## Data recorded after 8pm is attached to the next day
sleep_data$day           <- date(sleep_data$datetime_from + hours(4))
sleep_data$duration      <- int_length(
                              sleep_data$datetime_from %--% sleep_data$datetime_to
                            ) / period_to_seconds(minutes(1))
sleep_data$bed_time      <- case_when(
                              sleep_data$RAW_KIND == 1 ~ sleep_data$datetime_from
                            )
sleep_data$wakeup_time   <- case_when(
                              sleep_data$RAW_KIND == 1 ~ sleep_data$datetime_to
                            )

## Convert RAW_KIND to the corresponding activity and summarize values
sleep_data <- sleep_data %>%
  spread(RAW_KIND,duration,sep="_") %>%
  group_by(day) %>%
  summarise(
    sleep       = sum(RAW_KIND_1,  na.rm = TRUE),
    deep_sleep  = sum(RAW_KIND_2,  na.rm = TRUE),
    nap         = sum(RAW_KIND_3,  na.rm = TRUE),
    deep_nap    = sum(RAW_KIND_4,  na.rm = TRUE),
    walk        = sum(RAW_KIND_5,  na.rm = TRUE),
    run         = sum(RAW_KIND_6,  na.rm = TRUE), # Really not sure about this one
    bed_time    = min(bed_time,    na.rm = TRUE),
    wakeup_time = max(wakeup_time, na.rm = TRUE)
  )
```

# Visualisation
## Steps
### Distribution of steps
```{r}
data <- steps_data %>%
  mutate(date = date(datetime)) %>%
  select(date,STEPS) %>%
  group_by(date) %>%
  summarise(steps = sum(STEPS))

ggplot(data, aes(steps)) +
  geom_histogram(binwidth = 500) +
  theme_minimal() +
  labs(x="Steps", y="Number of occurences")
```

### Distribution of steps per day of week
```{r}
data <- steps_data %>%
  mutate(
    date = date(datetime),
    wday = wday(
      datetime,
      label = TRUE,
      week_start = getOption("lubridate.week.start", 1)
    )
  ) %>%
  select(date, wday, STEPS) %>%
  group_by(date, wday) %>%
  summarise(steps = sum(STEPS))
  
ggplot(data, aes(x=wday,y=steps)) +
  geom_boxplot() +
  theme_minimal() +
  labs(x="Day of week", y="Number of steps")
```

### Number of steps per month
```{r}
data <- steps_data %>%
  mutate(month = floor_date(datetime, unit = "month")) %>%
  select(month,STEPS) %>%
  group_by(month) %>%
  summarise(steps = sum(STEPS))

ggplot(data, aes(x=month, y=steps)) +
  geom_col() +
  scale_x_datetime(labels = date_format("%Y-%m")) + 
  scale_y_continuous(
    breaks = seq(0,500000,50000),
    labels=function(x) format(x, big.mark = " ")
  ) + 
  geom_smooth(method = lm) + 
  theme_minimal() + 
  labs(x="Month", y="Number of steps")
```

### Average number of steps per hour of the day, year after year
```{r}
data <- steps_data %>%
  mutate(
    date = date(datetime),
    time = datetime-floor_date(datetime, unit="day"),
    year = year(datetime)
  ) %>%
  group_by(date) %>%
  mutate(cumsteps = cumsum(STEPS)) %>%
  select(date, time, year, cumsteps) %>%
  ungroup() %>%
  group_by(time, year) %>%
  summarise(min = min(cumsteps), max = max(cumsteps), average = mean(cumsteps))

ggplot(data) + 
  geom_step(aes(x=time, y=average)) + 
  theme_minimal() + 
  labs(title = "Average number of steps per hour of the day", x="Hour", y="Number of steps") +
  scale_x_continuous(
    breaks = seq(
      0,
      period_to_seconds(hours(24)),
      period_to_seconds(hours(1))
    ),
    labels = seq(0,24,1)
  ) + 
  facet_wrap(vars(year))

```

### Active time per week
```{r}
data <- sleep_data %>%
  mutate(
    wday = wday(
      day,
      label = TRUE,
      week_start = getOption("lubridate.week.start", 1)
    )
  ) %>%
  select(day, wday, walk) %>%
  group_by(day, wday) %>%
  summarise(walk_time = sum(walk)/60)
  
ggplot(data, aes(x=wday,y=walk_time)) +
  geom_boxplot() +
  theme_minimal() +
  labs(x="Day of week", y="Hours active")
```

## Sleep
### Distribution of sleep duration
```{r}
data <- sleep_data %>%
  group_by(day) %>%
  summarise(sleep_duration = sum(sleep)/60)

ggplot(data) +
  geom_histogram(aes(sleep_duration), bins = 50) +
  scale_x_continuous(breaks = seq(0,12,1)) +
  theme_minimal() +
  labs(x="Sleep duration (hours)", y="Number of occurences")
```

### Distribution of deep sleep duration
```{r}
data <- sleep_data %>%
  group_by(day) %>%
  summarise(deep_sleep_duration = sum(deep_sleep)/60)

ggplot(data) +
  geom_histogram(aes(deep_sleep_duration),bins = 50) +
  scale_x_continuous(breaks = seq(0,12,1)) +
  theme_minimal() +
  labs(x="Deep sleep duration (hours)", y="Number of occurences")
```

### Distribution of sleep duration per day of week, year after year
```{r}
data <- sleep_data %>%
  mutate(
    wday = wday(
      day,
      label = TRUE,
      week_start = getOption("lubridate.week.start", 1)
    ),
    year = year(day)
  ) %>%
  select(year, day, wday, sleep) %>%
  group_by(year, day, wday) %>%
  summarise(sleep_duration = sum(sleep)/60)
  
ggplot(data, aes(x=wday,y=sleep_duration)) +
  geom_boxplot() +
  theme_minimal() +
  labs(x="Day of week", y="Sleep duration") +
  facet_grid(rows = vars(year), )
```

### Distribution of nap duration
```{r}
data <- sleep_data %>%
  filter(nap > 0) %>%
  group_by(day) %>%
  summarise(nap_time = sum(nap))

ggplot(data) +
  geom_histogram(aes(nap_time), bins = 10) +
  scale_x_continuous(breaks = seq(0,240,15)) +
  theme_minimal() +
  labs(x="Nap duration (minutes)", y="Number of occurences")
```

### Time of bed and waking up by year
```{r}
data <- sleep_data %>%
  mutate(
    year = year(day),
    month = floor_date(day, unit = "month"),
    bed_time_hms = hms::as.hms(
                                period_to_seconds(
                                  hours(hour(bed_time)) + minutes(minute(bed_time))
                                )
                              ),
    wakeup_time_hms = hms::as.hms(
                                   period_to_seconds(
                                     hours(hour(wakeup_time))+minutes(minute(wakeup_time))
                                   )
                                 )
  ) %>%
  drop_na(bed_time_hms) %>%
  drop_na(wakeup_time_hms)

ggplot(data) +
  geom_histogram(aes(bed_time_hms), fill="orange", alpha=0.5, bins=30) +
  geom_histogram(aes(wakeup_time_hms), fill="blue", alpha=0.5, bins=30) +
  scale_x_continuous(
    breaks = seq(
      0,
      period_to_seconds(hours(24)),
      period_to_seconds(hours(1))
    ),
    labels = seq(0,24,1)
  ) +
  theme_minimal() +
  labs(x="Bed time and wakeup time", y="Number of occurences") +
  facet_grid(rows = vars(year),scales="free_y")
```