1 Project Description
Tidy Tuesday has a weekly data project aimed at the R ecosystem. An emphasis is placed on understanding how to summarize and arrange data to make meaningful charts with ggplot2, tidyr, dplyr, and other tools in the tidyverse ecosystem.
2 Dataset
Data came from R4DS. The data includes the date, information about the number of members, number of messages, and the breakdown by direct messages, public, private, or shared channels.
3 Setup
3.1 Load Libraries
if (!require("pacman")) install.packages("pacman")
pacman::p_load("tidyverse","visdat","grid","gridExtra")
theme_set(theme_minimal())3.2 Import Data
df <- read_csv("https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2019/2019-07-16/r4ds_members.csv")4 Exploratory Data Analysis
- There are 678 observations and 21 variables.
- name, guests, and messages_in_shared_channels only have one unique value in the column.
- total_membership and full_members are duplicate columns.
- entries prior to 2017-09-01 only have 1 member. This is likely when R4DS was just getting started.
4.1 View Data
glimpse(df)## Observations: 678
## Variables: 21
## $ date <date> 2017-08-27, 2017-08-28, ...
## $ total_membership <dbl> 1, 1, 1, 1, 1, 188, 284, ...
## $ full_members <dbl> 1, 1, 1, 1, 1, 188, 284, ...
## $ guests <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0...
## $ daily_active_members <dbl> 1, 1, 1, 1, 1, 169, 225, ...
## $ daily_members_posting_messages <dbl> 1, 0, 1, 0, 1, 111, 110, ...
## $ weekly_active_members <dbl> 1, 1, 1, 1, 1, 169, 270, ...
## $ weekly_members_posting_messages <dbl> 1, 1, 1, 1, 1, 111, 183, ...
## $ messages_in_public_channels <dbl> 4, 0, 0, 0, 1, 252, 326, ...
## $ messages_in_private_channels <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0...
## $ messages_in_shared_channels <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0...
## $ messages_in_d_ms <dbl> 1, 0, 0, 0, 0, 119, 46, 7...
## $ percent_of_messages_public_channels <dbl> 0.8000, 0.0000, 0.0000, 0...
## $ percent_of_messages_private_channels <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0...
## $ percent_of_messages_d_ms <dbl> 0.2000, 0.0000, 0.0000, 0...
## $ percent_of_views_public_channels <dbl> 0.2857, 1.0000, 1.0000, 1...
## $ percent_of_views_private_channels <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0...
## $ percent_of_views_d_ms <dbl> 0.7143, 0.0000, 0.0000, 0...
## $ name <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0...
## $ public_channels_single_workspace <dbl> 10, 10, 11, 11, 12, 12, 1...
## $ messages_posted <dbl> 35, 35, 37, 38, 66, 1101,...head(df)## # A tibble: 6 x 21
## date total_membership full_members guests daily_active_me~
## <date> <dbl> <dbl> <dbl> <dbl>
## 1 2017-08-27 1 1 0 1
## 2 2017-08-28 1 1 0 1
## 3 2017-08-29 1 1 0 1
## 4 2017-08-30 1 1 0 1
## 5 2017-08-31 1 1 0 1
## 6 2017-09-01 188 188 0 169
## # ... with 16 more variables: daily_members_posting_messages <dbl>,
## # weekly_active_members <dbl>, weekly_members_posting_messages <dbl>,
## # messages_in_public_channels <dbl>, messages_in_private_channels <dbl>,
## # messages_in_shared_channels <dbl>, messages_in_d_ms <dbl>,
## # percent_of_messages_public_channels <dbl>,
## # percent_of_messages_private_channels <dbl>,
## # percent_of_messages_d_ms <dbl>,
## # percent_of_views_public_channels <dbl>,
## # percent_of_views_private_channels <dbl>, percent_of_views_d_ms <dbl>,
## # name <dbl>, public_channels_single_workspace <dbl>,
## # messages_posted <dbl>summary(df)## date total_membership full_members guests
## Min. :2017-08-27 Min. : 1.0 Min. : 1.0 Min. :0
## 1st Qu.:2018-02-12 1st Qu.: 978.2 1st Qu.: 978.2 1st Qu.:0
## Median :2018-07-31 Median :1605.0 Median :1605.0 Median :0
## Mean :2018-07-31 Mean :1567.8 Mean :1567.8 Mean :0
## 3rd Qu.:2019-01-16 3rd Qu.:2142.8 3rd Qu.:2142.8 3rd Qu.:0
## Max. :2019-07-05 Max. :3029.0 Max. :3029.0 Max. :0
## daily_active_members daily_members_posting_messages weekly_active_members
## Min. : 1.00 Min. : 0.00 Min. : 1.0
## 1st Qu.: 63.00 1st Qu.: 6.00 1st Qu.:206.0
## Median : 88.00 Median : 11.00 Median :239.0
## Mean : 91.39 Mean : 13.24 Mean :249.7
## 3rd Qu.:110.00 3rd Qu.: 16.00 3rd Qu.:307.8
## Max. :258.00 Max. :111.00 Max. :525.0
## weekly_members_posting_messages messages_in_public_channels
## Min. : 1.00 Min. : 0.00
## 1st Qu.: 35.00 1st Qu.: 9.25
## Median : 48.00 Median : 19.00
## Mean : 52.16 Mean : 28.46
## 3rd Qu.: 59.00 3rd Qu.: 35.00
## Max. :278.00 Max. :326.00
## messages_in_private_channels messages_in_shared_channels messages_in_d_ms
## Min. : 0.000 Min. :0 Min. : 0.00
## 1st Qu.: 0.000 1st Qu.:0 1st Qu.: 1.00
## Median : 0.000 Median :0 Median : 4.00
## Mean : 1.718 Mean :0 Mean : 13.05
## 3rd Qu.: 0.000 3rd Qu.:0 3rd Qu.: 12.00
## Max. :75.000 Max. :0 Max. :227.00
## percent_of_messages_public_channels percent_of_messages_private_channels
## Min. :0.0000 Min. :0.0000
## 1st Qu.:0.5840 1st Qu.:0.0000
## Median :0.8000 Median :0.0000
## Mean :0.7248 Mean :0.0305
## 3rd Qu.:0.9444 3rd Qu.:0.0000
## Max. :1.0000 Max. :1.0000
## percent_of_messages_d_ms percent_of_views_public_channels
## Min. :0.0000 Min. :0.2726
## 1st Qu.:0.0345 1st Qu.:0.9115
## Median :0.1595 Median :0.9519
## Mean :0.2270 Mean :0.9285
## 3rd Qu.:0.3478 3rd Qu.:0.9744
## Max. :1.0000 Max. :1.0000
## percent_of_views_private_channels percent_of_views_d_ms name
## Min. :0.000000 Min. :0.00000 Min. :0
## 1st Qu.:0.000000 1st Qu.:0.02235 1st Qu.:0
## Median :0.000000 Median :0.04170 Median :0
## Mean :0.009773 Mean :0.06176 Mean :0
## 3rd Qu.:0.006450 3rd Qu.:0.07433 3rd Qu.:0
## Max. :0.267400 Max. :0.72170 Max. :0
## public_channels_single_workspace messages_posted
## Min. :10.0 Min. : 35
## 1st Qu.:15.0 1st Qu.:20543
## Median :19.0 Median :33828
## Mean :17.8 Mean :32936
## 3rd Qu.:21.0 3rd Qu.:40104
## Max. :27.0 Max. :59627sapply(df, function(x) n_distinct(x)) %>% sort()## guests messages_in_shared_channels
## 1 1
## name public_channels_single_workspace
## 1 18
## messages_in_private_channels daily_members_posting_messages
## 27 52
## messages_in_d_ms messages_in_public_channels
## 83 108
## weekly_members_posting_messages percent_of_messages_private_channels
## 114 120
## daily_active_members percent_of_views_private_channels
## 168 192
## weekly_active_members percent_of_messages_d_ms
## 267 323
## percent_of_messages_public_channels total_membership
## 330 443
## full_members percent_of_views_d_ms
## 443 489
## percent_of_views_public_channels messages_posted
## 520 674
## date
## 6784.2 Missing Values
There are no missing values in the data set.
#Visualize missing values
vis_miss(df, sort_miss=TRUE)
5 Data Wrangling
Remove/replace missing values and drop unnecessary columns
# remove columns with only 1 unique value
df <- df[, sapply(df, function(x) n_distinct(x)) > 1] %>% select(-full_members)
df <- df %>% filter(date >= "2017-09-01")6 Visualizations
- When R4DS launched, initial users were very active at first and then usage gradually declined.
- Numbers dramatically increased just before the start of 2018.
- What happened on 2018-06-19? Total messages posted dropped. Did the site go down?
# active vs total membership
df %>% ggplot(aes(x=date))+
geom_area(aes(y=total_membership, fill="Total Membership"))+
geom_area(aes(y=weekly_active_members, fill="Active Members"))+
scale_fill_manual(values = c("steelblue1","steelblue4")) +
labs(title="Active Members vs Total Membership", fill="")
# total members posting messages
df %>% ggplot(aes(x=date))+
geom_line(aes(y=weekly_members_posting_messages), color="red")+
labs(title="Number of Members Posting Messages")
# messages posted
df %>% ggplot(aes(x=date))+
geom_line(aes(y=messages_posted), color="purple4", size=1.2)+
labs(title="Total Messages Posted")
## FINAL VISUALIZATION
# active vs. total membership
p1 <- df %>% ggplot(aes(x=date))+
geom_area(aes(y=total_membership), fill="steelblue4")+
geom_area(aes(y=weekly_active_members), fill="steelblue1")+
labs(x="Date", y="Total Number")+
theme(axis.title = element_text(size=9))
# messages posted vs. membership
p2 <- df %>% ggplot(aes(x=date))+
geom_area(aes(y=total_membership, fill="Total Membership"))+
geom_area(aes(y=weekly_active_members, fill="Active Members"))+
geom_line(aes(y=messages_posted, color="Messages Posted"), size=1.2)+
scale_fill_manual(values = c("steelblue1","steelblue4")) +
scale_color_manual(values = "purple3") +
labs(x="Date", y="Total Number")+
theme(legend.position="bottom",
legend.title=element_blank(),
legend.spacing.x = unit(0.4, 'cm'),
axis.title = element_text(size=9))+
guides(fill = guide_legend(order=1),
color = guide_legend(order=2))
#https://github.com/hadley/ggplot2/wiki/Share-a-legend-between-two-ggplot2-graphs
#function to create a common legend for two plots
g_legend<-function(a.gplot){
tmp <- ggplot_gtable(ggplot_build(a.gplot))
leg <- which(sapply(tmp$grobs, function(x) x$name) == "guide-box")
legend <- tmp$grobs[[leg]]
return(legend)}
mylegend<-g_legend(p2)
tg <- grobTree(textGrob("Membership Status vs. Messages Posted",
y=1,
vjust=1,
gp=gpar(fontface="bold", fontsize = 16)),
textGrob("Messages posted are rapidly increasing, despite active members staying fairly constant",
y=0,
vjust=0,
gp = gpar(fontsize=11, col="grey20")),
cl="titlegrob")
heightDetails.titlegrob <- function(x) do.call(sum,lapply(x$children, grobHeight))
#create final plot
p3 <- grid.arrange(arrangeGrob(p1 + theme(legend.position="none"),
p2 + theme(legend.position="none"),
nrow=1),
mylegend,
nrow=2,
heights=c(10, 1),
top = tg)
ggsave("R4DS-community-stats.png", p3)