In hs97/kane2017: Cultural Origin and Academic Achievement

require(tidyverse) #you may need to run install.packages('tidyverse') first, only once on a machine
require(dplyr)
require(lubridate) #same goes for any new package
require(rvest)
require(stringr)
require(knitr)
require(ggplot2)
load("../data/honors.RData")
load(file = "../data/athletic_summary.RData")
require(gridExtra)
load("../data/students.RData")
load("../data/roster.RData")
load("../data/classified.RData")
load(file = "../data/ethnic_summary.RData")

knitr::opts_chunk$set(tidy=TRUE, message=FALSE)

Motivation and Approaches

• Motivation

  • Informal Response to the article : What Does it Mean to be the Best?: An Alum Considers the Relative Importance of Admission Criteria

  • Personal Interest in seeing how students of Asian origins are performing academically

• Approach

  • Attain information from graduation catalog and team rosters

  • Use the wru package to identify someone's ethnicity through surname analysis

  • Compare the percentage of students attaining cum laude among different demographics.

---

Disclaimers and Limitations

• Limitations of the Bayesian surname analysis. Inability to account for names such as "Lee"
• Loss of information during transition. For example: a student named Andrew might be recorded as Andy on the team roster
• Inability to account for walk-ons.
• Graduation honors are not the only metrics for a student's importance and contribution to the college!

---

Data

• Utilizes two datasets: student and roster
• The students dataset comes from the Williams College catalog archive, containing graduation data from the class of 2003 to the class of 2015

• Has r nrow(students) entries

  r head(students, 3) %>% mutate(name = c('*Amy Eph, with highest honor in Economics', '*Chris Somebody', '*Lauren Nobody')) %>% kable()

• The roster dataset comes forom Williams College athletic archives, including rosters from the class of 2002 to the class of 2015 and teams such as soccer, football, and cross country

• Has r nrow(roster) entries r head(roster, 3) %>% mutate(name = c('John Lax', 'Ryan Tennis', 'Will Football')) %>% kable()

---

Data Tidying and Manuvering

• Join the two initial dataframes
• Read honors based on the names
• Analyze ethnicities based on surnames
• summarize ratio for each subgroup for each year

• r kable(head(ethnic_summary, 3))

• A graphical summary of the data

  r p1 <- ethnic_summary %>% ggplot(aes(x = year, y = ratio, color = race)) + geom_point() + theme_bw() p2 <- honors %>% mutate(latin = ifelse(summa == 1 | magna == 1 | cum == 1, 1, 0)) %>% group_by(year, team) %>% summarize(ratio = sum(latin) / length(name)) %>% filter(team == "Women's Cross Country" | team == "Men's Soccer" | team == "football" | is.na(team)) %>% ggplot(aes(x = year, y = ratio, color = team)) + geom_point() + theme_bw() grid.arrange(p1, p2)

---

Two proportion z-test

-

hispanic <- classified %>%
  filter(hispanic == 1) %>%
  group_by(latin) %>%
  summarize(value = length(name))
non_hispanic = classified %>%
  filter(hispanic == 0) %>%
  group_by(latin) %>%
  summarize(value = length(name))
prop.test(c(filter(hispanic, latin == 1)$value, 
            filter(non_hispanic, latin == 1)$value), 
          c(sum(hispanic$value), sum(non_hispanic$value)))

• Independence
• Success/Failure Condition
• 10% Condition

---

T tests

athlete <- athletic_summary %>%
  filter(athlete == 1) 
non_athlete <- athletic_summary %>%
  filter(athlete == 0) 
t.test(athlete$ratio, non_athlete$ratio)

• Conditions
• Nearly Normal: r p1 <- ggplot(athlete, aes(x = ratio, fill = "athlete")) + geom_histogram(bins = 6, aes(y = ..density..)) + theme_bw() + stat_function(fun = dnorm, args = c(mean = mean(athlete$ratio), sd = sd(athlete$ratio))) p2 <- ggplot(non_athlete, aes(x = ratio, fill = "non-athlete")) + geom_histogram(bins = 6, aes(y = ..density..)) + theme_bw() + stat_function(fun = dnorm, args = c(mean = mean(non_athlete$ratio), sd = sd(non_athlete$ratio))) grid.arrange(p1, p2, nrow = 1)
• Independence

---

Thoughts and Potential Future Work

• Different subgroups of students might have different focuses
• Interesting data for the Davis center
• A foundation for a study on the relationship between socioeconomic status and thriving academically at college
• Compare results across colleges to examine if our institution is doing its best to support students of various demographics

hs97/kane2017 documentation built on May 17, 2019, 5:55 p.m.