1 Fitbit Time-Series Data

*describe how the data were collected, stored, and what they look like*
All the data that I’m using can be retrieved by calling the fitbit API through the getActivitiesResourceByDatePeriod, getHeartByDateIntraday, and getWeightByDate methods. The functions I wrote to perform these GET requests can be found in this script.

I decided to structure the data in a tidy format. The code to collect new data from the fitbit servers and tidy it can be found in this script. So first I split up the data in minute-level and daily-level time series. Then I created separate tibbles per type of value:

• Minute-level
  • Numerical: calories, distance, elevation, floors, heartrate, steps
  • Ordinal: activity intensity
• Daily-level
  • Numerical: weight, BMI, activity calories, base metabolic heartrate calories, rest heartrate{, minutes, calories out (per heartrate zone –> separate table) [backlog]}

str(fitbit_data)

## List of 3
##  $ dl_num:Classes 'tbl_df', 'tbl' and 'data.frame':  611 obs. of  5 variables:
##   ..$ date : Date[1:611], format: "2019-01-01" ...
##   ..$ type : Factor w/ 5 levels "bmi","weight",..: 3 4 5 3 4 5 3 4 5 3 ...
##   ..$ value: num [1:611] 1885 708 59 1885 1389 ...
##   ..$ day  : Ord.factor w/ 7 levels "Monday"<"Tuesday"<..: 2 2 2 3 3 3 4 4 4 5 ...
##   ..$ week : Ord.factor w/ 21 levels "1"<"2"<"3"<"4"<..: 1 1 1 1 1 1 1 1 1 1 ...
##  $ ml_ord:Classes 'tbl_df', 'tbl' and 'data.frame':  202980 obs. of  5 variables:
##   ..$ datetime: POSIXct[1:202980], format: "2019-01-01 00:00:00" ...
##   ..$ type    : Factor w/ 1 level "intensity": 1 1 1 1 1 1 1 1 1 1 ...
##   ..$ value   : Ord.factor w/ 4 levels "sedentary"<"light"<..: 1 2 1 1 1 1 1 2 2 2 ...
##   ..$ day     : Ord.factor w/ 7 levels "Monday"<"Tuesday"<..: 2 2 2 2 2 2 2 2 2 2 ...
##   ..$ week    : Ord.factor w/ 21 levels "1"<"2"<"3"<"4"<..: 1 1 1 1 1 1 1 1 1 1 ...
##  $ ml_num:Classes 'tbl_df', 'tbl' and 'data.frame':  1193695 obs. of  5 variables:
##   ..$ datetime: POSIXct[1:1193695], format: "2019-01-01 00:00:00" ...
##   ..$ type    : Factor w/ 6 levels "calories","distance",..: 1 2 3 4 5 6 1 2 3 4 ...
##   ..$ value   : num [1:1193695] 1.7 0 0 0 80 ...
##   ..$ day     : Ord.factor w/ 7 levels "Monday"<"Tuesday"<..: 2 2 2 2 2 2 2 2 2 2 ...
##   ..$ week    : Ord.factor w/ 21 levels "1"<"2"<"3"<"4"<..: 1 1 1 1 1 1 1 1 1 1 ...

1.1 Minute-Level Data

1.1.1 Data Validation

*Show summary statistics for all numerical variables and table for ordinal. Show histogram and density plot for heartrate. Other plots are so heavily skewed that they are not very informative other than that they show that the values are 0 most of the time.*

fitbit_data$ml_num %>%
  group_by(type) %>%
  summarise(
  minimum = min(value),
  pctl_25 = quantile(value, probs = c(0.25), names = FALSE),
  median = median(value),
  pctl_75 = quantile(value, probs = c(0.75), names = FALSE),
  maximum = max(value),
  mean = mean(value),
  st_dev = sd(value)
  ) %>%
  knitr::kable(digits = 2, caption = "", col.names = c("", 
                                                       "*P~0~*",
                                                       "*P~25~*",
                                                       "*P~50~*",
                                                       "*P~75~*",
                                                       "*P~100~*",
                                                       "*$\\mu$*",
                                                       "*$\\sigma$~X~*"))

	P0	P25	P50	P75	P100	\(\mu\)	\(\sigma\)X
calories	1.22	1.26	1.36	1.66	17.94	2.35	2.35
distance	0.00	0.00	0.00	0.00	249.47	9.22	26.72
elevation	0.00	0.00	0.00	0.00	21.34	0.07	0.61
floors	0.00	0.00	0.00	0.00	7.00	0.02	0.20
heartrate	35.00	50.00	58.00	69.00	182.00	62.69	19.94
steps	0.00	0.00	0.00	0.00	185.00	11.31	30.20

*Some text here.*

min_hr <- fitbit_data$ml_num %>%
  filter(type == "heartrate") %>%
  select(value) %>%
  min()

max_hr <- fitbit_data$ml_num %>%
  filter(type == "heartrate") %>%
  select(value) %>%
  max()

fitbit_data$ml_num %>%
  filter(type == "heartrate") %>%
  ggplot(aes(x = value, fill = value, stat(density))) +
  geom_histogram(binwidth = 1, color = "black", fill = "#2A211C") +
  geom_density(alpha = 0.8, color = "black", fill = "#9F2042") +
  theme_classic(base_size = 12, base_line_size = 1) +
  theme(axis.title.x = element_text(color = "#000000", face = "italic", margin = margin(7.5, 0, 0, 0)),
        axis.title.y = element_text(color = "#000000", face = "italic", margin = margin(0, 7.5, 0, 0)),
        axis.text.x = element_text(color = "#000000", face = "bold", size = 9),
        axis.text.y = element_text(color = "#000000", face = "bold", size = 9),
        panel.background = element_rect(fill = "#FCFCFC"),
        plot.background = element_rect(color = "#000000", fill = "#FCFCFC")) +
  labs(x = "Heart rate", y = "Density") +
  scale_x_continuous(breaks = c(min_hr, 50, 75, 100, 125, 150, 175, max_hr))

ds_histogram_and_density_heartrate

*Some text here.*

# act_tidy %>% 
#   filter(!is.na(act_tidy$datetime)) %>%
#   group_by(as.Date(datetime, tz = "Europe/Amsterdam"), type) %>%
#   summarise(
#     daily_total = sum(value)
#   ) %>%
#   group_by(type) %>%
#   summarise(
#     total = sum(daily_total),
#     minimum = min(daily_total),
#     mean = mean(daily_total),
#     median = median(daily_total),
#     maximum = max(daily_total)
#   ) %>% 
#   knitr::kable(digits = 2, caption = "Summary of activity data")

# act_tidy %>% 
#   filter(!is.na(act_tidy$datetime)) %>%
#   group_by(week, type) %>%
#   summarise(
#     average_day = sum(value)/uniqueN(day)
#   ) %>%
#   ggplot(aes(x = week, y = average_day)) +
#   geom_point() +
#   geom_line() +
#   facet_wrap(~ type, nrow = 2, ncol = 3, scales = "free_y")

2 How does getting fitter reflect in the data?

*obviously do a solid attempt to answer this question here*

# hr_summary %>%
#   ggplot(aes(x = date, y = rest_hr)) +
#   geom_point() +
#   geom_smooth(se = FALSE, span = 0.275) +
#   labs(title = "Rest Heart Rate over Time",
#        subtitle = "*this is how you make a subtitle, Luc*")

# hr_intraday %>%  
#   group_by(week) %>%
#   summarise(
#     minimum = min(hr),
#     mean = mean(hr),
#     median = median(hr),
#     maximum = max(hr)
#   ) %>% 
#   gather(key = "type", value = "value", -week) %>%
#   ggplot(aes(x = week, y = value)) +
#   geom_point() +
#   geom_line() +
#   facet_wrap(~ type, nrow = 2, ncol = 2, scales = "free_y")

# hr_intraday %>%
#   mutate(week = fct_rev(as.factor(week))) %>%
#   ggplot(aes(x= hr, y = week)) +
#   geom_density_ridges(scale = 2.5, alpha = 0.7) +
#   xlim(30, 194) +
#   theme_ridges()

3 Useful Resources

• Fitbit Web API Basics
• Fitbit Web API Swagger UI
• Strava Web API Swagger UI