Mattermost Lunch Channel History

• Data Source: Mattermost API, CCTB instance
• Tasks:
  • Part I - June 2024: retrieving chat history data through the mattermost API
  • Part II - September 2024: analyzing messages in the lunch channel
  • Part III - September 2024: specific tasks
• Language: R

Select one of the following tasks

General comment: your estimate in step 1 does not need to be perfect, settle for a heuristic that is good enough

Task A - most crowded day of the week

• estimate the total number of people having lunch (or coffee) at the CCTB/mensa for each day → when was the time that most people went to lunch?
• plot the number of people per day over time (also try to summarize by week/month/year)
• plot a boxplot for the number of people per day of the week → what is the most crowded day of the week?
• make the same plot as above, but separately for every month/year → is there a shift in day of the week preference?
• perform a statistical test for the hypothesis: "Mondays and Fridays are less crowded than Tuesday and Thursday"
• discuss caveats of the data and methods used

Task B - lunch time

• estimate the time of lunch/coffee for each day → when was the most popular time?
  • try to consider proposed times ("mensa at 12?", "11:15?")
  • direct calls ("mensa?", "now")
  • relative times ("lunch in 5min", "mensa in half an hour?")
• plot the lunch time over the years (also try to summarize by week/month/year) → is there a trend (gradual shift or break point(s)) in lunch time?
• plot a boxplot for the lunch time per day of the week → is there a difference in lunch time per day of the week?
• make the same plot as above, but separately for every month/year → is the pattern above consistent over the year(s)?
• perform a statistical test for the hypothesis: "Lunch time is later during semester break (April,May,August,September) than during lecture period since 2022"
• discuss caveats of the data and methods used

Task C - your own idea

If you have other ideas, feel free to follow them, but create a plan similar to that for Task A and B above, before you start.

library(tidyverse)

── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
✔ dplyr     1.1.4     ✔ readr     2.1.5
✔ forcats   1.0.0     ✔ stringr   1.5.1
✔ ggplot2   3.5.0     ✔ tibble    3.2.1
✔ lubridate 1.9.3     ✔ tidyr     1.3.1
✔ purrr     1.0.2     

── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
✖ purrr::%||%()   masks base::%||%()
✖ dplyr::filter() masks stats::filter()
✖ dplyr::lag()    masks stats::lag()
ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors

theme_set(theme_light())

set.seed(42)
"Sabine Jana Sascha Mike Markus" %>%  str_split(" ") %>% unlist %>% sample %>% str_c(collapse=" → ")

'Sabine → Markus → Mike → Sascha → Jana'

1. Data loading

Load files:

• messages.csv
• reactions.csv
• files.csv

messages <- read_csv("messages.csv")
reactions <- read_csv("reactions.csv")
files <- read_csv("files.csv")

Rows: 6947 Columns: 6

── Column specification ────────────────────────────────────────────────────────
Delimiter: ","
chr  (3): post_id, message, username
dbl  (2): num_reactions, num_files
dttm (1): create_at

ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

Rows: 3634 Columns: 5

── Column specification ────────────────────────────────────────────────────────
Delimiter: ","
chr  (4): post_id, emoji_name, username, emoji
dttm (1): create_at

ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

Rows: 221 Columns: 3

── Column specification ────────────────────────────────────────────────────────
Delimiter: ","
chr (3): post_id, file_id, link

ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

Task A

As result, we want a table of the form | day of the week | number of people | |------------------------|--------------------------| | Monday | 10 | | Tuesday | 11 |

messages %>% head

A tibble: 6 × 6

post_id	create_at	message	username	num_reactions	num_files
<chr>	<dttm>	<chr>	<chr>	<dbl>	<dbl>
h957oi1t63rf5b6p4pqfhiis5h	2016-05-02 12:02:13	Also ich habe etwas dabei	greatfireball	0	0
rpwuii8hci8ezmn3d41u9d58fc	2016-05-02 12:03:02	ich gehe mir jetzt was holen	nterhoeven	0	0
e8bnwq8c7fdf5k7zpqyeihougr	2016-05-02 12:03:24	Alles klar	greatfireball	0	0
sku34dhpntdfdmc5d9ky9a9yso	2016-05-02 12:20:33	Hallo @joerg	greatfireball	0	0
hi38ssjpb3yqz88dk83tzjq5zr	2016-05-02 12:20:35	Essen?	greatfireball	0	0
dux74xuq738u38z384wg8mri1e	2016-05-02 12:20:36	:smile:	greatfireball	0	0

reactions %>% head

A tibble: 6 × 5

post_id	create_at	emoji_name	username	emoji
<chr>	<dttm>	<chr>	<chr>	<chr>
ng3m9o5c5igw3gpgg1dqmgs4mw	2017-08-18 11:57:23	+1	lulei	👍
atnyf76f1tb1fmznhrk4s3wohr	2017-08-21 11:50:04	+1	sonja	👍
atnyf76f1tb1fmznhrk4s3wohr	2017-08-21 11:52:22	+1	iimog	👍
eh5ge6nnf7np5q61rqpfwhmqaw	2017-08-23 11:24:45	+1	sonja	👍
eh5ge6nnf7np5q61rqpfwhmqaw	2017-08-23 11:26:44	+1	alexanderkeller	👍
eh5ge6nnf7np5q61rqpfwhmqaw	2017-08-23 11:49:19	+1	t.paul	👍

reaction_users_per_day <- reactions %>%
    filter(year(create_at) >= 2022) %>%
    mutate(day = floor_date(create_at, "day")) %>%
    select(day, username) %>% 
    unique()

message_users_per_day <- messages %>%
    filter(year(create_at) >= 2022) %>%
    mutate(day = floor_date(create_at, "day")) %>%
    select(day, username) %>% 
    unique()

message_users_per_day %>% head

A tibble: 6 × 2

day	username
<dttm>	<chr>
2022-01-04	iimog
2022-01-04	art
2022-01-04	andreaskuhn
2022-01-05	andreaskuhn
2022-01-10	simon.schardt
2022-01-11	jscabral

users_per_day <- full_join(reaction_users_per_day,message_users_per_day,by=c("day","username"))

library(lubridate)

users_per_day %>%
    mutate(weekday = wday(day, label=TRUE)) %>%
    count(weekday) %>%
    ggplot(aes(x=weekday, y=n)) + geom_col()

counts_per_wday <- users_per_day %>%
    count(day) %>%
    mutate(weekday = wday(day, label=TRUE))

counts_per_wday %>%
    ggplot(aes(x=weekday, y=n)) + geom_boxplot()

counts_per_wday %>%
    filter(weekday %in% c("Mon", "Tue", "Thu", "Fri")) %>%
    ggplot(aes(x=n)) + geom_histogram() + facet_wrap(~weekday)

`stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

mean_var_per_wday <- counts_per_wday %>% 
    filter(weekday %in% c("Mon", "Tue", "Thu", "Fri")) %>%
    group_by(weekday) %>%
    summarise(median= median(n),mean = mean(n), var = var(n), n = n())

mean_var_per_wday

A tibble: 4 × 5

weekday	median	mean	var	n
<ord>	<dbl>	<dbl>	<dbl>	<int>
Mon	5	5.144000	4.059742	125
Tue	5	4.907692	4.223971	130
Thu	4	4.644068	3.479067	118
Fri	4	4.371901	4.585537	121

monday <- counts_per_wday %>%
    filter(weekday == "Mon")
tuesday <- counts_per_wday %>%
    filter(weekday == "Tue")
thursday <- counts_per_wday %>%
    filter(weekday == "Thu")
friday <- counts_per_wday %>%
    filter(weekday == "Fri")

Our Bonferroni-corrected significance level:

0.05/6

0.00833333333333333

library(stats)
wilcox.test(monday$n, tuesday$n, alternative = "two.sided")

	Wilcoxon rank sum test with continuity correction

data:  monday$n and tuesday$n
W = 8667.5, p-value = 0.3514
alternative hypothesis: true location shift is not equal to 0

wilcox.test(thursday$n, friday$n, alternative = "two.sided")

	Wilcoxon rank sum test with continuity correction

data:  thursday$n and friday$n
W = 7797.5, p-value = 0.2124
alternative hypothesis: true location shift is not equal to 0

wilcox.test(monday$n, friday$n, alternative = "two.sided")

	Wilcoxon rank sum test with continuity correction

data:  monday$n and friday$n
W = 9229, p-value = 0.002527
alternative hypothesis: true location shift is not equal to 0

wilcox.test(tuesday$n, friday$n, alternative = "two.sided")

	Wilcoxon rank sum test with continuity correction

data:  tuesday$n and friday$n
W = 9069.5, p-value = 0.03412
alternative hypothesis: true location shift is not equal to 0

wilcox.test(monday$n, thursday$n, alternative = "two.sided")

	Wilcoxon rank sum test with continuity correction

data:  monday$n and thursday$n
W = 8427.5, p-value = 0.05149
alternative hypothesis: true location shift is not equal to 0

wilcox.test(tuesday$n, thursday$n, alternative = "two.sided")

	Wilcoxon rank sum test with continuity correction

data:  tuesday$n and thursday$n
W = 8237, p-value = 0.3089
alternative hypothesis: true location shift is not equal to 0