# Analysis of hours spent on Master's degree

## Introduction

In this post I will show how to analyze data from toggl to visualzation the time that I spent on my Master’s degree. I studied mathematical statistics at Gothenburg University from September 2018 to June 2020. I kept track of my hours using toggl. Unfortunately I did not keep track of the time spent at my lectures as I thought I could pull the data from a calendar later. So all of the time estimates are based on my outside of class studying. I generally study using the Pomodoro Technique with a timer for 25 minutes. I would also stop the timer immediately if I checked my phone or got up for anything. This is clear from the data as there are lots of entries under 25 minutes. We’ll first go over extracting the data from toggl and then some visualizations.

### Setup

library(togglr)
library(tidyr)
library(lubridate)
library(dplyr)
library(knitr)
library(ggplot2)
library(scales)
library(knitr)

## Extracting Data from Toggl

First we pull the data from toggl using the R package togglr. I have each one of my courses as a separate project on toggl so first I create a tribble of the meta data for each course. Each course is 50% time and taking two classes a quarter (four a semester) is full time. The abbriviation HT is Hösttermin which means fall semester and VT = Vårtermin which means spring semester. Each of the semesters are divided into two parts, 1 and 2. So an entire year consists of the reading periods HT1, HT2, VT1, VT2 spanning about 10 weeks. Now we will dive in to extracting and processing the data from toggl.

courses <- tribble(
~course, ~short_name, ~semester, ~period, ~study_year,
"Statistical Learning", "Statistical Learning", "VT", 2, 1,
"Time Series", "Time Series", "VT", 2, 1,
"Foundations of probability theory", "Probability Theory", "HT", 2, 1,
"Bayesian Stats Chalmers", "Bayesian Stats", "HT", 1, 1,
"Linear Statistical Models", "Linear Models", "HT", 2, 1,
"Linear Mixed Models for Longitudinal Data", "Linear Mixed Models", "VT", 1, 1,
"Machine Learning GU", "Machine Learning (NLP)", "HT", 1, 1,
"Experimental design and sampling", "Experimental Design", "VT", 1, 1,
"Time Series", "Time Series", "VT", 2, 1,
"Perspectives in Mathematics", "Math History", "HT", 2, 1,
"Integration Theory", "Integration Theory", "HT", 1, 2,
"Functional Analysis", "Functional Analysis", "HT", 2, 2,
"Topology", "Topology", "VT", 1, 2,
"Master's Thesis", "Thesis", "", NA, 2
)

schedule <- tribble(
~semester, ~period, ~study_year, ~int,
"HT", 1, 1, as.POSIXct("2018-09-03", tz = "CET") %--% as.POSIXct("2018-11-05", tz = "CET"),
"HT", 2, 1, as.POSIXct("2018-11-05", tz = "CET") %--% as.POSIXct("2019-01-20", tz = "CET"),
"VT", 1, 1, as.POSIXct("2019-01-20", tz = "CET") %--% as.POSIXct("2019-03-23", tz = "CET"),
"VT", 2, 1, as.POSIXct("2019-03-25", tz = "CET") %--% as.POSIXct("2019-06-20", tz = "CET"),
"HT", 1, 2, as.POSIXct("2019-09-02", tz = "CET") %--% as.POSIXct("2019-11-04", tz = "CET"),
"HT", 2, 2, as.POSIXct("2019-11-04", tz = "CET") %--% as.POSIXct("2020-01-18", tz = "CET"),
"VT", 1, 2, as.POSIXct("2020-01-20", tz = "CET") %--% as.POSIXct("2020-03-21", tz = "CET"),
"VT", 2, 2, as.POSIXct("2020-03-23", tz = "CET") %--% as.POSIXct("2020-06-10", tz = "CET"),
)

# Add an id to each row
period_ids <- 1:nrow(schedule)
schedule$period_key <- period_ids # Get the time entries by month since I have a lot of them start_day <- date("2018-08-28") last_day <- date("2020-06-14") # Divide the days into 8 intervals # Toggl api only returns 1000 entries so need to partition date range seg <- 8 time_interval <- seq(start_day, last_day, length.out = seg) since = time_interval[-seg] until = time_interval[-1] # Use the toggl api to get the time entries time_entries_list <- lapply(2:seg, function(i) { togglr::get_time_entries(since = time_interval[i - 1], until = time_interval[i]) }) # Use rbind.fill to create columns if they do not exist toggl_df <- do.call(plyr::rbind.fill, time_entries_list) %>% dplyr::rename(course = project_name) toggl_joined <- inner_join(toggl_df, courses, by = "course") toggl_simple <- select(toggl_joined, start, stop, short_name, description, duration, semester, period, study_year) %>% mutate(start = with_tz(start, tzone = "CET"), stop = with_tz(stop, tzone = "CET")) %>% dplyr::rename(course = short_name) toggl_simple$period_key <- unlist(lapply(toggl_simple$start, function(s) { cond = s %within% schedule$int
matches = sum(cond)
if(matches == 0) {
NA
} else if(matches > 1) {
warning("Found multiple matches")
period_ids[cond]
} else {
period_ids[cond]
}
}))

day_order_abbr <- c("Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun")

time_entries <- left_join(toggl_simple, schedule, by = "period_key") %>%
mutate(
semester = if_else(is.na(semester.x) | semester.x == "", semester.y, semester.x),
period = if_else(is.na(period.x), period.y, period.x),
study_year = if_else(is.na(study_year.x), study_year.y, study_year.x),
) %>%
select(start, stop, course, description, duration, semester, period, study_year) %>%
# Remove all duplicates
distinct(start, stop, .keep_all = T) %>%
mutate(date = format(start, "%Y-%m-%d"),
weekday = factor(weekdays(start, abbreviate = T), levels = day_order_abbr))

Now we have a nicely formatted data frame of the time entries for all my course entries. Next we want to summarize the total time by each day.

### Daily Totals

The naïve way creating daily totals would be to just group by date. One issue with this approach is that it ignores the dates in which I didn’t do any studying. I want any date that is in a quarter to be in the data set but with an NA or 0 for the daily total. We do this below by creating date sequences for each quarter. Then we join this to the original time entries to fill in the gaps of study days.

# All days in school
md <- mapply(seq, from = int_start(schedule$int), to = int_end(schedule$int), by = 'day')

# All the days in each semester with meta data
md_meta <- lapply(seq_along(md), function(i) {
cbind(
data.frame(date = format(md[[i]], "%Y-%m-%d"),
weekday = factor(weekdays(md[[i]], abbreviate = T), levels = day_order_abbr),
stringsAsFactors = FALSE),
schedule[rep(i, length(md[[i]])),]
)
})

school_days <- do.call("rbind", md_meta)

# Fill the missing days with 0
daily_entries <- time_entries %>%
group_by(date) %>%
# Ensure we use dplyr summarize instead of plyr
dplyr::summarize(total_hours = sum(duration) / 3600) %>%
right_join(school_days, by = "date") %>%
replace_na(list(total_hours = 0))

## Basic summary statistics

Now we can create a data frame with some basic summary statistics from the data. Don’t worry about the content of these yet as we will be plotting them later.

sum_stats <- time_entries %>%
unite(quarter, semester, period, sep = "") %>%
group_by(course, quarter, study_year) %>%
summarize(
n = n(),
total_time_hours = sum(duration) / 3600,
mean_time_minutes = mean(duration) / 60,
)

# Add the total hours spent on each course
# Only useful for thesis that spans multiple reading periods
sum_stats_course <- sum_stats %>%
group_by(course, study_year) %>%
summarize(
total_time_course = sum(total_time_hours)
)

sum_stats <- left_join(sum_stats, sum_stats_course, by = "course") %>%
arrange(course, quarter)

sum_stats$cumsum <- sum_stats$total_time_hours
thesis_cond <- sum_stats$course == "Thesis" # Update total hours in thesis to cumsum sum_stats[thesis_cond, "cumsum"] <- cumsum(sum_stats[thesis_cond,"total_time_hours"]) ### Time spent per year I spent about 77 more hours in my second year than the first. It seems like that should be even higher but that was probably due to more work on personal projects. I didn’t record those hours as rigorously as my course work entries but it would be interesting to look at those as well. time_entries %>% group_by(study_year) %>% summarize(total_hours = sum(duration) / 3600) %>% kable(digits = 1, col.names = c("Study Year", "Total Hours")) Study Year Total Hours 1 484.3 2 560.7 ## Visualizations First set some colors and theme # CB Palette like ESL cbPalette <- c( "#999999", "#E69F00", "#56B4E9", "#009E73", "#F0E442", "#0072B2", "#D55E00", "#CC79A7") theme_set( theme_minimal(base_size = 15, base_family = "Arial") %+replace% theme( plot.caption = element_text(size = 9, color = "gray50", hjust = 1) ) ) # Per course ggplot(sum_stats) + geom_bar( aes(x = reorder(course, total_time_course), y = total_time_hours, fill=quarter), stat="identity", position = position_stack(reverse = TRUE)) + geom_text( aes(x = course, y = cumsum, label = round(cumsum)), hjust = 1, nudge_y = -1, size = 6 ) + coord_flip() + labs( title = "Hours spent on each course by quarter", caption = "stefanengineering.com", x = NULL, y = "Total Time (Hours)" ) + theme( panel.grid.major = element_blank(), panel.grid.minor = element_blank(), legend.position = c(0.85,0.25), legend.box.background = element_rect(colour = "black") ) + scale_fill_manual(values = cbPalette[-1]) Figure 1: Total hours spent on each course by quarter In Figure 2 we see the total hours that I spent on each course, colored by year. The second year courses were all math courses compared with the first year when I took all statistics courses. I took integration theory without any other courses and spent much more time than other courses. Similarly for functional analysis I focused essentially 100% of my time on it. Overall both of these ended up being more than double the time I would spent on a statistic course which was about 58 hours in median. Interestingly, I spent much less time on topology which is the other math course I took. This was due to the fact that I had to learn topology for functional analysis any way so it was much easier as a result. # Per course ggplot(sum_stats_course) + geom_bar( aes(x = reorder(course, total_time_course), y = total_time_course, fill=as.factor(study_year)), stat="identity") + geom_text( aes(x = course, y = total_time_course, label = round(total_time_course)), hjust = 1, nudge_y = -1, size = 6 ) + coord_flip() + labs( title = "Hours spent on each course by year", caption = "stefanengineering.com", x = NULL, y = "Total Time (Hours)" ) + theme( panel.grid.major = element_blank(), panel.grid.minor = element_blank(), legend.position = c(0.85,0.25), legend.box.background = element_rect(colour = "black") ) + scale_fill_manual(values = cbPalette[-1], name = "Year") Figure 2: Total hours spent on each course by year Looking at the data by quarter was the most interesting for me. It is clear from this that I spent more time in every quarter during the second year. Also, the first quarter of spring is massively much less time than the other quarters. Especially the first year but the second year as well I had a really hard time adjusting to the swedish winters. time_entries %>% group_by(study_year, semester, period) %>% summarize(hours = sum(duration)/3600) %>% unite(quarter, semester, period, sep = "") %>% ggplot(.) + geom_bar(aes(x = quarter, y = hours, fill = as.factor(study_year)), position = "dodge", stat = "identity") + scale_fill_manual(values = cbPalette[-1], name = "Year") + theme(panel.grid.major.x = element_blank(), panel.grid.minor.x = element_blank()) + labs(title = "Hours per quarter", caption = "stefanengineering.com") ### Calendar Using calendarPlot from the openair package we can display the number of hours per day on a calendar. The days with 0 hours spent studying are marked as grey while the white squares are not part of that quarter. max_time <- max(daily_entries$total_hours)

daily_entries %>%
group_by(study_year, semester, period) %>%
mutate(date = date(date)) %>%
group_walk(~ openair::calendarPlot(.x,
pollutant = "total_hours", year = c(2018, 2019, 2020),
w.shift = 2,
# Force all plots to have the same limits
limits = c(0, max_time),
# Set the 0 to be grey so we can see the missed days
cols = c("#DCDCDC", tail(openair::openColours(), n=50)),
main = paste0("Year ", .y["study_year"], " ", .y["semester"], .y["period"]))) %>%
invisible()        ### Study streaks

After looking at the calendar graphs I wanted to see what the longest streaks I had for studying and for breaks. I never tried to keep a streak alive at any point but I did end up with quite a few long ones. My longest streak was 32 days of work and my longest break was 11 days when I went home for Easter break.

run_df <- daily_entries %>%
mutate(studied = total_hours > 0) %>%
group_by(study_year, semester, period) %>%
# Compute the start/end days for runs
# Modified from https://stackoverflow.com/a/46961802/1351718
group_map(~with(rle(.x$studied), data.frame(is_study = values, length = lengths, start = .x$date[cumsum(lengths) - lengths + 1],
end = .x$date[cumsum(lengths)], # Add the group variables .y )[order(-lengths),])) toprun_lst <- lapply(run_df, function(x) { s <- x[x$is_study,]
b <- x[!x$is_study,] rbind( s[which.max(s$length),],
b[which.max(b$length),] ) }) toprun_df <- do.call("rbind", toprun_lst) row.names(toprun_df) <- NULL toprun_df %>% arrange(study_year, semester, period) %>% mutate(quarter = paste0("Y", study_year, " ", semester, period)) %>% select(is_study, length, quarter) %>% group_by(is_study) %>% group_walk(~print(ggplot(.x) + geom_bar(aes(x = quarter, y = length), fill = cbPalette, stat="identity") + ggtitle(ifelse(.y, "Study streak", "Days off streak")) + theme( axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1), panel.grid.major.x = element_blank(), panel.grid.minor.x = element_blank() ) ) ) %>% invisible()  ### When was I studying? I studied at approximately the same times each day. I am very much a morning person and tend to do my best work after waking up and before lunch. Usually I would wake up around 6:00 and video chat with my girlfriend and then study if I did not have school. To create this plot I create a sequence of times for each day, with 10 minutes spacing. Then for each of these times I compute how many time intervals it falls into (summing all the indicator functions). This is like manually computing a histogram with taking into account that the data are intervals. time_seq <- function(dates, entries, by = "1 min") { ds <- seq(dates, dates + days(1), by = by) res <- NULL # Alternatively could use group_by, group_walk/group_map for(i in seq_along(dates)) { d <- dates[i] # Set the date. This prevents the sequences from having different lengths due to leap date(ds) <- d d_entries <- filter(entries, date == d) # Compute overlap of time entries for one date r <- unlist(lapply(ds, function(td) { sum(td >= d_entries$start & td <= d_entries$stop) })) # Add to the result if(is.null(res)) res <- r else res <- res + r } list(count = res, date = ds) } ti <- 10 time_df <- data.frame( time_seq(dates = unique(floor_date(time_entries$start, unit = "day")),
time_entries,
by = paste(ti, "min")))
ggplot(time_df) +
geom_bar(aes(x = date, y = count), stat = "identity", width = ti * 60) +
scale_x_datetime(breaks = "3 hour", labels=date_format("%H:%M")) +
ylab("count") +
xlab("time") I did most of my studying around 9:00 and 13:00. It seems as if this is a Gaussian mixture of four normal distributions centered at 6:00, 9:00, 13:00 and 18:30. Unlike most students, I hardly did any studying in the evening or at night.

### Distribution of event time

What was the distribution of the time for each study period? As I mentioned in the introduction, I worked in 25 minute blocks for most of the time. Somewhat frequently I would get distracted and stop the time early leading to many events shorter than 25 minutes. It is easy to the effect on the number of events close to 25 minutes. The large outliers came from some of the projects that I had to do. I find the Pomodoro Technique works really well when studying math or statistics theory but not as well for programming or data analysis. It seems from the histogram that the data could be exponential or some power law mixed with a point mass around 25 minutes. This is kind of like a zero inflated Poisson model, except something like a 25-inflated Poisson model.

ggplot(data.frame(minutes = time_entries\$duration / 60)) +
geom_histogram(aes(x = minutes), bins = 50) ## Conclusion

Overall it was quite interesting to look at how I studied during my Master’s degree. I still keep track of the time I spend on my activities. Next up I want to link this to some of my other personal data such as rock climbing frequency, running and weight. Other sources that I could join together are weather data and sunrise/sunset data.

## Reproducibility

─ Session info ───────────────────────────────────────────────────────────────────────────────────────────────────────
setting  value
version  R version 3.6.3 (2020-02-29)
os       macOS Mojave 10.14.6
system   x86_64, darwin15.6.0
ui       X11
language (EN)
collate  en_US.UTF-8
ctype    en_US.UTF-8
tz       Europe/Stockholm
date     2020-06-19

─ Packages ───────────────────────────────────────────────────────────────────────────────────────────────────────────
package       * version  date       lib source
assertthat      0.2.1    2019-03-21  CRAN (R 3.6.0)
bibtex          0.4.2    2017-06-30  CRAN (R 3.6.0)
blogdown        0.13     2019-06-11  CRAN (R 3.6.0)
bookdown        0.11     2019-05-28  CRAN (R 3.6.0)
cli             2.0.2    2020-02-28  CRAN (R 3.6.0)
cluster         2.1.0    2019-06-19  CRAN (R 3.6.3)
codetools       0.2-16   2018-12-24  CRAN (R 3.6.3)
colorspace      1.4-1    2019-03-18  CRAN (R 3.6.0)
crayon          1.3.4    2017-09-16  CRAN (R 3.6.0)
curl            4.3      2019-12-02  CRAN (R 3.6.0)
digest          0.6.25   2020-02-23  CRAN (R 3.6.0)
dplyr         * 0.8.5    2020-03-07  CRAN (R 3.6.0)
ellipsis        0.3.0    2019-09-20  CRAN (R 3.6.0)
evaluate        0.14     2019-05-28  CRAN (R 3.6.0)
fansi           0.4.1    2020-01-08  CRAN (R 3.6.0)
farver          2.0.3    2020-01-16  CRAN (R 3.6.0)
getPass         0.2-2    2017-07-21  CRAN (R 3.6.0)
ggplot2       * 3.3.1    2020-05-28  CRAN (R 3.6.2)
glue            1.4.1    2020-05-13  CRAN (R 3.6.2)
gtable          0.3.0    2019-03-25  CRAN (R 3.6.0)
hexbin          1.27.3   2019-05-14  CRAN (R 3.6.0)
highr           0.8      2019-03-20  CRAN (R 3.6.0)
hms             0.5.3    2020-01-08  CRAN (R 3.6.0)
htmltools       0.4.0    2019-10-04  CRAN (R 3.6.0)
httr            1.4.1    2019-08-05  CRAN (R 3.6.0)
jsonlite        1.6.1    2020-02-02  CRAN (R 3.6.0)
keyring         1.1.0    2018-07-16  CRAN (R 3.6.0)
knitcitations * 1.0.8    2017-07-04  CRAN (R 3.6.0)
knitr         * 1.28     2020-02-06  CRAN (R 3.6.0)
labeling        0.3      2014-08-23  CRAN (R 3.6.0)
lattice         0.20-38  2018-11-04  CRAN (R 3.6.3)
latticeExtra    0.6-28   2016-02-09  CRAN (R 3.6.0)
lifecycle       0.2.0    2020-03-06  CRAN (R 3.6.0)
lubridate     * 1.7.4    2018-04-11  CRAN (R 3.6.0)
magrittr        1.5      2014-11-22  CRAN (R 3.6.0)
mapproj         1.2.7    2020-02-03  CRAN (R 3.6.0)
maps            3.3.0    2018-04-03  CRAN (R 3.6.0)
MASS            7.3-51.5 2019-12-20  CRAN (R 3.6.3)
Matrix          1.2-18   2019-11-27  CRAN (R 3.6.3)
mgcv            1.8-31   2019-11-09  CRAN (R 3.6.3)
munsell         0.5.0    2018-06-12  CRAN (R 3.6.0)
nlme            3.1-144  2020-02-06  CRAN (R 3.6.3)
openair         2.7-2    2020-04-02  CRAN (R 3.6.2)
parsedate       1.2.0    2019-05-08  CRAN (R 3.6.0)
pillar          1.4.3    2019-12-20  CRAN (R 3.6.0)
pkgconfig       2.0.3    2019-09-22  CRAN (R 3.6.0)
plyr            1.8.6    2020-03-03  CRAN (R 3.6.0)
prettyunits     1.1.1    2020-01-24  CRAN (R 3.6.0)
purrr           0.3.3    2019-10-18  CRAN (R 3.6.0)
R6              2.4.1    2019-11-12  CRAN (R 3.6.0)
RColorBrewer    1.1-2    2014-12-07  CRAN (R 3.6.0)
Rcpp            1.0.3    2019-11-08  CRAN (R 3.6.0)
readr           1.3.1    2018-12-21  CRAN (R 3.6.0)
RefManageR      1.2.12   2019-04-03  CRAN (R 3.6.0)
rematch2        2.0.1    2017-06-20  CRAN (R 3.6.0)
rlang           0.4.5    2020-03-01  CRAN (R 3.6.0)
rmarkdown       2.1      2020-01-20  CRAN (R 3.6.0)
rstudioapi      0.11     2020-02-07  CRAN (R 3.6.0)
scales        * 1.1.0    2019-11-18  CRAN (R 3.6.0)
sessioninfo   * 1.1.1    2018-11-05  CRAN (R 3.6.0)
stringi         1.4.6    2020-02-17  CRAN (R 3.6.0)
stringr         1.4.0    2019-02-10  CRAN (R 3.6.0)
tibble          2.1.3    2019-06-06  CRAN (R 3.6.0)
tidyr         * 1.0.2    2020-01-24  CRAN (R 3.6.0)
tidyselect      1.0.0    2020-01-27  CRAN (R 3.6.0)
togglr        * 0.1.34   2020-06-12  Github (ThinkR-open/togglr@ea3afde)
vctrs           0.2.4    2020-03-10  CRAN (R 3.6.0)
withr           2.1.2    2018-03-15  CRAN (R 3.6.0)
xfun            0.12     2020-01-13  CRAN (R 3.6.0)
xml2            1.2.5    2020-03-11  CRAN (R 3.6.0)
yaml            2.2.1    2020-02-01  CRAN (R 3.6.0)

 /Library/Frameworks/R.framework/Versions/3.6/Resources/library