Tidyverse Create - Extended
Coco Donovan, Kory Martin
2023-04-29
Introduction:
For my tidyverse create assignment, I chose a data set containing roster information for all NCAA Women’s Basketball teams. I intend to use readr to read in my data, dplyr to manipulate my data and ggplot2 to display my analysis.
Loading/Installing Packages
library(ggplot2)
library(dplyr)##
## Attaching package: 'dplyr'## The following objects are masked from 'package:stats':
##
## filter, lag## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, unionlibrary(readr)readr
read_csv()
I used the read_csv() function to read in my csv of NCAA Women’s Basketball roster info. Within read_csv() I used the col_types argument to change data types of certain columns. Initially the columns that I changed inside read_csv() were chr type, but to do meaningful analysis I needed to make them numerical, hence setting them to col_double() (which sets the data type of the column to a double). This same logic can be applied to height_
ncaa_wbb_rosters <- read_csv('https://raw.githubusercontent.com/Sports-Roster-Data/womens-college-basketball/main/wbb_rosters_2022_23.csv', col_types = list(total_inches = col_double(), height_ft = col_double(), height_in = col_double()))
ncaa_avg_height <- round(mean(ncaa_wbb_rosters$total_inches, na.rm = TRUE),2)dyplr
distinct()
I am unsure what values the redshirt column could possibly take on, so I use a pipe and the distinct() function to highlight the possible values ‘redshirt’ may take on. I find that redshirt can either be 1, for “yes, a student athlete was redshirted,” or 0, for “no, a student athlete was not redshirted.”
knitr::kable(ncaa_wbb_rosters %>%
distinct(redshirt))| redshirt |
|---|
| 0 |
| 1 |
select(), group_by(), summarize(), and arrange()
Now, I wanted to get a glimpse of the teams with the tallest average height. To do this I used pipes. I selected the team variable and the total_inches variable. Then I grouped by height using the group_by function. I then used the summarize function to provide the counts of players for players per team the corresponding average height per team.
avg_heights <- ncaa_wbb_rosters %>%
select('team','total_inches') %>%
group_by(team) %>%
summarize(number_of_players = n(), Avg_height = round(mean(total_inches),2)) %>%
arrange(desc(Avg_height))
knitr::kable(head(avg_heights))| team | number_of_players | Avg_height |
|---|---|---|
| Indiana | 13 | 73.38 |
| UConn | 12 | 73.25 |
| Michigan | 15 | 73.07 |
| South Carolina | 13 | 72.85 |
| Rutgers | 8 | 72.75 |
| Stanford | 15 | 72.73 |
ggplot2
Plotting Average Heights per team
The primary package I used here was ggplot2, although I do start this code chunk off with some dplyr. I used top_n() so that I could get the stop six results for the teams with the highest average heights. Then I piped those top 6 results into a bar chart. I made sure to change the order of the x-axis as the default is to sort the items alphabetically, but I want the value to be sorted based height in a descending order. I then limited the x-axis to better range to show the difference between the top average heights (using coord_cartesian()), and finished off my visualization by displaying the numerical values on top of each bar using geom_text() and displaying a line to show the average height of an NCAA women’s basketball team for perspective.
top_n(avg_heights, n=6, Avg_height) %>%
ggplot(., mapping = aes(x=reorder(team, desc(Avg_height)),y=Avg_height, fill=team)) +
geom_bar(stat='identity') +
coord_cartesian(ylim = c(66,74)) +
geom_hline(aes(yintercept=ncaa_avg_height,linetype = "NCAA Average Height")) +
geom_text(aes(label = Avg_height), vjust = 1.5,
position = position_dodge(width = 0.9))+
xlab("Team") +
ylab("Average Height (in)")
Extending the Examples (by Kory Martin):
For this we will extend upon the examples presented, by looking at the number of international students that are attending the universities and gleaning some meaningful insights based on this.
We begin by using the mutate function in dplpyr to create a new column that will hold a value of ‘international’ if the student is not from the US, and a value of ‘domestic’ if they are from the US. In addition to the mutate function, we will use the if_else function to assign the value to the new column, based on whether or not the value of country_clean is ‘USA’ or not.
(ncaa_wbb_roster_extended <- ncaa_wbb_rosters %>%
mutate(domestic_international = ifelse(country_clean == 'USA', 'domestic', 'international')))## # A tibble: 13,806 × 30
## ncaa_id team player_id name year hometown homestate high_school
## <dbl> <chr> <dbl> <chr> <chr> <chr> <chr> <chr>
## 1 721 Air Force 11807 Mackenzie Le Fres… Elk Gro… Californ… St. Francis
## 2 721 Air Force 11805 Milahnie Pe… Fres… Tampa Florida Seffner Ch…
## 3 721 Air Force 11801 Madison Smi… Soph… Connell Washingt… Connell
## 4 721 Air Force 11795 Taylor Britt Juni… Columbia South Ca… Spring Val…
## 5 721 Air Force 11796 Kamri Heath Seni… Edmond Oklahoma Edmond Nor…
## 6 721 Air Force 11800 Kayla Pilson Juni… Houston Texas Westside
## 7 721 Air Force 11806 Faith Shelt… Fres… Stockton Californ… Lincoln
## 8 721 Air Force 11804 Griffin Gre… Fres… Monument Colorado Lewis-Palm…
## 9 721 Air Force 11803 Parker Brown Fres… Spokane Washingt… Mead
## 10 721 Air Force 11798 Dasha Macmi… Juni… Colleyv… Texas Grapevine
## # ℹ 13,796 more rows
## # ℹ 22 more variables: previous_school_clean <chr>, height_clean <chr>,
## # position <chr>, jersey <chr>, url <chr>, season <chr>, team_state <chr>,
## # conference <chr>, division <chr>, height_ft <dbl>, height_in <dbl>,
## # total_inches <dbl>, primary_position <chr>, secondary_position <chr>,
## # position_clean <chr>, year_clean <chr>, redshirt <dbl>, hs_clean <chr>,
## # hometown_clean <chr>, state_clean <chr>, country_clean <chr>, …We will then use select function combined with the slice_sample function to confirm that our new column has the expected values, by selecting a random sample of rows and looking at their country_clean and domestic_international values
ncaa_wbb_roster_extended %>% select(country_clean, domestic_international) %>% slice_sample(n=15)## # A tibble: 15 × 2
## country_clean domestic_international
## <chr> <chr>
## 1 USA domestic
## 2 USA domestic
## 3 USA domestic
## 4 USA domestic
## 5 USA domestic
## 6 USA domestic
## 7 USA domestic
## 8 USA domestic
## 9 USA domestic
## 10 USA domestic
## 11 USA domestic
## 12 USA domestic
## 13 USA domestic
## 14 USA domestic
## 15 USA domesticNext we will use the group_by function to generate a count of the number of domestic vs international students at each college. Furthermore, we will use the summarize function to create summary measures for the total number of domestic students, the total number of international students, and the total number of players on the team. Finally, based on these counts, we will then create an additional variable pct_international that will represent the pct of international students on the student rosters.
(ncaa_international_domestic <- ncaa_wbb_roster_extended %>%
group_by(team) %>%
summarize(num_domestic = sum(ifelse(domestic_international == 'domestic',1,0)),
num_international = sum(ifelse(domestic_international == 'international',1,0)),
num_players = n()) %>%
mutate(pct_international = num_international/num_players))## # A tibble: 938 × 5
## team num_domestic num_international num_players pct_international
## <chr> <dbl> <dbl> <int> <dbl>
## 1 A&M-Corpus Chri… 10 5 15 0.333
## 2 Abilene Christi… 13 0 13 0
## 3 Academy of Art 6 2 8 0.25
## 4 Adams St. 14 0 14 0
## 5 Adelphi 14 0 14 0
## 6 Air Force 13 0 13 0
## 7 Akron 10 2 12 0.167
## 8 Alabama 13 0 13 0
## 9 Alabama A&M 15 0 15 0
## 10 Alabama State 14 0 14 0
## # ℹ 928 more rowsWe will use the left_join function to connect our data to the division in which the college belongs to
(ncaa_international_domestic <- left_join(ncaa_international_domestic, ncaa_wbb_rosters %>% select(team, division) %>% distinct(), by=c("team"="team")))## # A tibble: 938 × 6
## team num_domestic num_international num_players pct_international division
## <chr> <dbl> <dbl> <int> <dbl> <chr>
## 1 A&M-Co… 10 5 15 0.333 I
## 2 Abilen… 13 0 13 0 I
## 3 Academ… 6 2 8 0.25 II
## 4 Adams … 14 0 14 0 II
## 5 Adelphi 14 0 14 0 II
## 6 Air Fo… 13 0 13 0 I
## 7 Akron 10 2 12 0.167 I
## 8 Alabama 13 0 13 0 I
## 9 Alabam… 15 0 15 0 I
## 10 Alabam… 14 0 14 0 I
## # ℹ 928 more rowsFinally we will use the group_by and summarize functions, to determine the average pct of international students per team, across teams in each of the three divisions.
ncaa_international_domestic %>%
group_by(division) %>%
summarize(num_teams = n(),
avg_pct_international = mean(pct_international))## # A tibble: 3 × 3
## division num_teams avg_pct_international
## <chr> <int> <dbl>
## 1 I 360 0.152
## 2 II 245 0.0735
## 3 III 333 0.0110Conclusion:
As you can see Indiana has the tallest average height, followed by UConn, Michigan, South Carolina, Rutgers and Stanford. As a huge women’s basketball fan, one thing that stands out to me is that all but two of these teams have been nationally ranked in the top 5 this season, and of the two that have not been ranked in the top 5, Rutgers and Michigan, Michigan has been ranked in the top 20, pretty consistently. Rutgers has been going through a rough patch in the absence of their hall-of-fame coach C. Vivian Stringer and their roster only contains 8 players (their average height may be inflated by the lack of players on the roster). However, my first thought after looking at this would be that it seems that height plays some component in basketball (stating the obvious).