{"id":945,"date":"2024-09-02T09:44:17","date_gmt":"2024-09-02T13:44:17","guid":{"rendered":"https:\/\/www.clayford.net\/statistics\/?p=945"},"modified":"2024-09-02T09:46:01","modified_gmt":"2024-09-02T13:46:01","slug":"us-names-section-2-3-of-regression-and-other-stories","status":"publish","type":"post","link":"https:\/\/www.clayford.net\/statistics\/us-names-section-2-3-of-regression-and-other-stories\/","title":{"rendered":"US Names: section 2.3 of Regression and Other Stories"},"content":{"rendered":"

In section 2.3 of Regression and Other Stories, Gelman et al.\u00a0discuss using graphs to learn more about data. The third example involves the exploration of names in the United States from 1880 to 2010. In this post I recreate figures 2.6 – 2.9 using R code that makes more sense to me. The original data and code are in this GitHub repo<\/a>.<\/p>\n

First read in the data and drop the superfluous first column which just contains row numbers:<\/p>\n

allnames <- read.csv("https:\/\/github.com\/avehtari\/ROS-Examples\/raw\/master\/Names\/data\/allnames_clean.csv")\r\nallnames$X <- NULL\r\ndim(allnames)<\/code><\/pre>\n
## [1] 98012   133<\/code><\/pre>\n
Notice this data set is quite large and \u201cwide\u201d with over 98,000 rows and 133 columns. There is one row per name per sex with counts per year spanning from 1880 to 2010. Here\u2019s a portion of the data:<\/p>\n
allnames[1:3,1:6]<\/code><\/pre>\n
##   name sex X1880 X1881 X1882 X1883\r\n## 1 Mary   F  7065  6919  8149  8012\r\n## 2 Anna   F  2604  2698  3143  3306\r\n## 3 Emma   F  2003  2034  2303  2367<\/code><\/pre>\n
Apparently 7,065 girls were named Mary in 1880, and then 6,919 in 1881, and so on.<\/p>\n
Figs 2.6 and 2.7 show the distribution of the last letters of boys\u2019 first names for the years 1906, 1956, and 2006. I wrote a function to create the plots for a given year.<\/p>\n
last_letter <- function(year){\r\n  xyear <- paste0("X", year)\r\n  d <- subset(allnames, sex == "M", select = c("name", xyear))\r\n  n <- nchar(d$name)\r\n  d$last <- substr(d$name, n, n)\r\n  d$last <- factor(d$last, levels = letters, labels = letters)\r\n  p_data <- aggregate(d[[xyear]], by = list(d[["last"]]), sum)\r\n  names(p_data) <- c("letter", "count")\r\n  p_data$p <- p_data[["count"]]\/sum(p_data[["count"]])\r\n  barplot(p ~ letter, data = p_data, \r\n          main = paste("Last letter of boys' names in", year))\r\n}<\/code><\/pre>\n
Now create the plots:<\/p>\n
last_letter(1906)<\/code><\/pre>\n
<\/p>\n
last_letter(1956)<\/code><\/pre>\n
<\/p>\n
last_letter(2006)<\/code><\/pre>\n
<\/p>\n
The 2006 plot shows a trend in giving boys names that end with the letter \u201cn\u201d, such as Ethan, Jayden, Aiden, Mason, and Logan.<\/p>\n
Fig 2.8 looks at the distribution of the letters of boys\u2019 last names over time. To create this plot I found it helpful to use the {tidyr} and {dplyr} packages. First I reshaped the data to \u201clong\u201d format using pivot_longer()<\/code> so there is one row per name per sex per year. Then I removed the \u201cX\u201d from the year value and once again extracted the last letter of the last names using the nchar()<\/code> and substr()<\/code> functions.<\/p>\n
library(tidyr)\r\nallnames_long <- pivot_longer(allnames, cols = X1880:X2010, \r\n                              names_to = "year", values_to = "count")\r\nallnames_long$year <- sub("X", "", allnames_long$year)\r\nn <- nchar(allnames_long$name)\r\nallnames_long$last <- substr(allnames_long$name, n, n)\r\nallnames_long$last <- factor(allnames_long$last, \r\n                             levels = letters, labels = letters)<\/code><\/pre>\n
Next I used some {dplyr} functions to calculate the proportions of boys\u2019 names ending in one of the 26 English letters by year. In the last step I create a new variable called \u201clast2\u201d to indicate if the letter was an N, D or Y. This is to help create a plot matching figure 2.8<\/p>\n
library(dplyr)\r\nlast_letter_males <- allnames_long |> \r\n  filter(sex == "M") |> \r\n  group_by(year, last) |> \r\n  summarise(letter_count = sum(count)) |> \r\n  mutate(p = letter_count\/sum(letter_count),\r\n         last2 = case_when(last == "n" ~ "n",\r\n                           last == "d" ~ "d",\r\n                           last == "y" ~ "y",\r\n                           .default = "other"))<\/code><\/pre>\n
Before I create the plot, I cut the data into two sets: one with letters N, D and Y; and one with the rest of the letters.<\/p>\n
ndy <- subset(last_letter_males, last2 != "other")\r\nother <- subset(last_letter_males, last2 == "other")<\/code><\/pre>\n
And now to create the plot using {ggplot2}. Using two data sets allows me to tweak the lines for the letters N, D and Y The \u201cSet2\u201d palette is a good all-purpose color blind friendly palette.<\/p>\n
library(ggplot2)\r\nggplot() +\r\n  aes(x = year, y = p, group = last) +\r\n  geom_line(data = other, alpha = 1\/5) +\r\n  geom_line(mapping = aes(color = last2), data = ndy, linewidth = 1.5) +\r\n  scale_x_discrete(breaks = c("1900", "1950", "2000")) +\r\n  scale_color_brewer(palette = "Set2") +\r\n  theme_minimal() +\r\n  labs(color = "letter")<\/code><\/pre>\n
<\/p>\n
Names ending in D and Y had their moments in the sun, but now it\u2019s all about names ending in N.<\/p>\n
Figure 2.9 displays trends in the concentrations of names in the top 10. This was the easiest plot to create. Again I used some {dplyr} reasoning to create the data. After I calculate proportions of counts by year and sex, I sort the data in descending order, slice off the top 10 names, and then sum the proportions by year and sex.<\/p>\n
names_conc <- allnames_long |>\r\n  group_by(year, sex) |> \r\n  mutate(p = count\/sum(count)) |> \r\n  arrange(desc(p)) |> \r\n  slice_head(n = 10) |> \r\n  summarize(total_p = sum(p))<\/code><\/pre>\n
And here\u2019s the plot:<\/p>\n
ggplot(names_conc) + \r\n  aes(x = year, y = total_p, group = sex, color = sex) +\r\n  geom_line() +\r\n  scale_x_discrete(breaks = c("1900", "1950", "2000")) +\r\n  scale_color_brewer(palette = "Set2") +\r\n  ylim(c(0, 0.45)) +\r\n  theme_minimal()<\/code><\/pre>\n
<\/p>\n
Pre 1900, about 40% of all boys\u2019 names were in the top 10. Fast forward 100 years to 2000, only about 10% of names, boy or girl, are in the top 10. We see much more variability in names in the 21st century.<\/p>\n","protected":false},"excerpt":{"rendered":"
In section 2.3 of Regression and Other Stories, Gelman et al.\u00a0discuss using graphs to learn more about data. The third… Read more<\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[13],"tags":[84],"class_list":["post-945","post","type-post","status-publish","format-standard","hentry","category-using-r","tag-regression-and-other-stories"],"_links":{"self":[{"href":"https:\/\/www.clayford.net\/statistics\/wp-json\/wp\/v2\/posts\/945","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.clayford.net\/statistics\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.clayford.net\/statistics\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.clayford.net\/statistics\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.clayford.net\/statistics\/wp-json\/wp\/v2\/comments?post=945"}],"version-history":[{"count":3,"href":"https:\/\/www.clayford.net\/statistics\/wp-json\/wp\/v2\/posts\/945\/revisions"}],"predecessor-version":[{"id":948,"href":"https:\/\/www.clayford.net\/statistics\/wp-json\/wp\/v2\/posts\/945\/revisions\/948"}],"wp:attachment":[{"href":"https:\/\/www.clayford.net\/statistics\/wp-json\/wp\/v2\/media?parent=945"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.clayford.net\/statistics\/wp-json\/wp\/v2\/categories?post=945"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.clayford.net\/statistics\/wp-json\/wp\/v2\/tags?post=945"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}