sample_data.Rmd

---
title: "Analysis of 24Hr Movement Guideline Fibit Data"
author: "Daniel Fuller"
date: "26/07/2021"
output:
  html_document:
    keep_md: true
---

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

# 24 Hour Movement Guidelines Data Analysis

## Required packages
```{r, echo = FALSE, include = FALSE}
library(tidyverse)
library(lubridate)
library(ggplot2)
library(naniar)
library(ggmap)
library(ggthemes)
library(sf)
library(rgeos)
library(cancensus)
library(cowplot)
library(ggspatial)
library(knitr)
library(ggpubr)
library(reshape)
library(scales)
```

# Data Description

There are a number of data types and frequencies. Below is a summary: 

1. Minute Frequency Data
    - heartrate_1min_merged
    - minuteCaloriesNarrow_merged
    - minuteIntensitiesNarrow_merged
    - minuteMETsNarrow_merged
    - minuteSleep_merged 
    - minuteStepsNarrow_merged
    
2. Daily Frequency Data
    - dailyCalories_merged
    - dailyIntensities_merged
    - dailySteps_merged
    - sleepDay_merged
    - heartRateZones_merged
        - Daily time in 4 heart rate zones (Out of Range, Fat Burn, Cardio, Peak)

3. Other Frequency data
    - activitylogs_merged 
        - An activity is recorded whenever the user inputs it
    - battery_merged
        - Battery level is recorded whenever there is a sync event
    - syncEvents_merged
        - All Sync events
    - 30secondSleepStages_merged
        - Sleep stages every 30 seconds. 

# To do

1. Combine the minute level data
2. Combine day level data
3. Aggregate minute level data to the day and confirm it makes sense with day level
4. Join battery and sync events to the minute level data
5. Aggregate everything up to the day level

## 1. Combine the minute level data

### Heart rate per minute
```{r}
hr_data <- read_csv("Data/heartrate_1min_merged.csv")
hr_data$time <- mdy_hms(hr_data$Time)
hr_data$Time <- NULL
colnames(hr_data) <- c("id", "heart_rate_bmp", "time")
summary(hr_data$heart_rate_bmp)
ggplot(hr_data, aes(heart_rate_bmp)) + 
        geom_histogram() 
```

### Calories per minute
```{r}
ee_data <- read_csv("Data/minuteCaloriesNarrow_merged.csv")
ee_data$time <- mdy_hms(ee_data$ActivityMinute)
ee_data$ActivityMinute <- NULL
colnames(ee_data) <- c("id", "calories", "time")
summary(ee_data$calories)
ggplot(ee_data, aes(calories)) + 
        geom_histogram() 
```

See fitbit description for calories [https://dev.fitbit.com/build/reference/web-api/intraday/get-activity-intraday-by-date/](https://dev.fitbit.com/build/reference/web-api/intraday/get-activity-intraday-by-date/)

```{}
## Using BMR/EER Algorithms

If a user had no Fitbit tracker data for the specific day then the greater of Logged Activities + BMR (for minutes when there is no activity) and the calories calculated from the EER for that day (if EER enabled for this user's profile) are taken. In case, there was some data from the tracker for the specific day, that data where available is used and for time where data is unavailable, the BMR is used. If the total is less than 20% greater than BMR then the EER (cals < EER * 0.8) is used. EER never used to calculate calories for today.
Using BMR Formula

Fitbit uses the standard MD Mifflin-St Jeor equation:

9.99 * weightKg + 6.25*heightCm - 4.92*ageYears + s, where s is +5 for 
males
and -161 for female

EER Formula (TEE total energy expenditure)

The EER Formula is based on http://www.cdc.gov/pcd/issues/2006/oct/pdf/06_0034.pdf, which in turn is based on "Food and Nutrition Board. Dietary reference intakes for energy, carbohydrate, fiber, fat, fatty acids, cholesterol, protein, and amino acids (macronutrients). Washington (DC): National Academy Press; 2005." http://www.nap.edu/openbook.php?isbn=0309085373&page=204
MALE-based EER Formula:

TEE = 864 - 9.72 x age (years) + 1.0 x (14.2 x weight(kg) + 503 x height
(meters))

FEMALE-based EER Formula:

TEE = 387 - 7.31 x age (years) + 1.0 x (10.9 x weight(kg) + 660.7 x height
(meters))
```

### Intensity per minute
```{r}
intense_data <- read_csv("Data/minuteIntensitiesNarrow_merged.csv")
intense_data$time <- mdy_hms(intense_data$ActivityMinute)
intense_data$ActivityMinute <- NULL
colnames(intense_data) <- c("id", "intensity", "time")
intense_data$intensity <- as.factor(intense_data$intensity)
table(intense_data$intensity)
ggplot(intense_data, aes(intensity)) + 
        geom_bar() 
```

This is a four category variable. Looks like it will match up with the heart rate zone data but not sure yet. 

### METS per minute
```{r}
met_data <- read_csv("Data/minuteMETsNarrow_merged.csv")
met_data$time <- mdy_hms(met_data$ActivityMinute)
met_data$ActivityMinute <- NULL
colnames(met_data) <- c("id", "mets", "time")
summary(met_data$mets)
ggplot(met_data, aes(mets)) + 
        geom_histogram() 
```

Interesting. Will need to look into how this is calculated. 10 METS as the median and 15 METS as the mean is not right. 

### Sleep per minute
```{r}
sleep_data <- read_csv("Data/minuteSleep_merged.csv")
sleep_data$time <- mdy_hms(sleep_data$date)
sleep_data$date <- NULL
colnames(sleep_data) <- c("id", "sleep", "log_id", "time")
sleep_data$sleep <- as.factor(sleep_data$sleep)
table(sleep_data$sleep)
ggplot(sleep_data, aes(sleep)) + 
        geom_bar() 
```

This is a three category variable. There is a lot of missing here so I'm guessing that these represent sleep stages or something with a null value meaning not sleeping.

### Steps per minute
```{r}
step_data <- read_csv("Data/minuteStepsNarrow_merged.csv")
step_data$time <- mdy_hms(step_data$ActivityMinute)
step_data$ActivityMinute <- NULL
colnames(step_data) <- c("id", "steps", "time")
summary(step_data$steps)
ggplot(step_data, aes(steps)) + 
        geom_histogram() 
```

## Joining the datasets

```{r}
ee_intense <- full_join(ee_data, intense_data, by = c("id", "time"))
ee_intense_met <- full_join(ee_intense, met_data, by = c("id", "time"))
ee_intense_met_step <- full_join(ee_intense_met, step_data, by = c("id", "time"))
ee_intense_met_step_hr <- full_join(ee_intense_met_step, hr_data, by = c("id", "time"))
data <- full_join(ee_intense_met_step_hr, sleep_data, by = c("id", "time"))

## Filter out coordinator data
data <- filter(data, id != "Study Coordinator QU")

### Create a location variable
data$location <- substr(data$id, 1, 1)

### Create an intervention 
data <- data %>%
              mutate(intervention = case_when(
                          time < "2021-10-25 00:00:00" ~ "pre",
                          time >= "2021-10-25 00:00:00" & time < "2021-12-05 00:00:00" ~ "intervention",
                          time > "2021-12-05 00:00:00" ~ "post"
                        ))
table(data$intervention)

### Recoding Sleep 
data <- data %>%
              mutate(sleep_yn = case_when(
                          sleep == 1 ~ "sleep",
                          sleep == 2 ~ "sleep",
                          sleep == 3 ~ "sleep",
                          TRUE ~ "awake"
                        ))
table(data$sleep_yn)


rm(ee_data, ee_intense, ee_intense_met, ee_intense_met_step, ee_intense_met_step_hr, hr_data, intense_data, met_data, sleep_data, step_data)
```

## Quick descriptive stats

```{r warning= FALSE}
data$day <- day(data$time)
data$hour <- hour(data$time)

hourly_summary <- data %>%
          group_by(id, hour) %>%
            get_summary_stats(calories, heart_rate_bmp, steps)

daily_summary <- data %>%
      group_by(id, day) %>%
        get_summary_stats(calories, heart_rate_bmp, steps)
```

## Heat Map over the study period per hour

```{r}
ggplot(hourly_summary, aes(x = hour, y = variable, fill = mean)) +
  geom_tile() + 
  facet_wrap(~ id) +
  scale_fill_gradient(low = "grey20", high = "red") +
  labs(x = "", y = "") +
  theme_classic() +
  theme(legend.position = "none") 
```

## Heat Map over the study period per day

```{r}
ggplot(daily_summary, aes(x = day, y = variable, fill = mean)) +
  geom_tile() + 
  facet_wrap(~ id) +
  scale_fill_gradient(low = "grey20", high = "red") +
  labs(x = "", y = "") +
  theme_classic() +
  theme(legend.position = "none") 
```

## 2. Combine day level data