Preliminary cleaning

Data were uploaded to GitHub for ease of access. Preliminary cleaning was accomplished with dplyr to create a dataset consisting of senators and representatives who’ve endorsed a candidate.

• read_csv reads the file from GitHub
• mutate recodes NAs in variable “endorsee” as “undeclared”
• select omits unnecessary columns
• filter filters the dataset to include only senators and representatives

rm(list=ls())
library(tidyverse)

#use read_csv from readr
endorse<-read_csv("https://raw.githubusercontent.com/sigmasigmaiota/elections/master/endorsements-2020.csv")

#replace NA with "undeclared" using mutate from dplyr
#remove unnecessary columns with select, from dplyr
#filter senators and representatives, from dplyr
senrep<-endorse%>%
  mutate(endorsee=ifelse(is.na(endorsee),"undeclared",endorsee))%>%
  select(position,endorsee)%>%
  filter(position=="senator"|position=="representative")

Barplot: geom_bar

qqplot2 requires a correctly tidied dataset, a geom, and a coordinate system; to function, aesthetic properties must be set for the geom. In the first plot below, aes is defined by the dataset senrep, with the fill command specifying that the color of each bar in the plot should correspond to each candidate (endorsee).

• geom_bar supplies the neccessary command for a bar plot
• facet_grid() creates multiple plots differentiated by position, or senator vs representative
• coord_flip() rotates the plots 90 degrees
• theme_bw() removes background color, which is gray by default
• theme(axis.text.x = element_text()) rotates labels on the x-axis 90 degrees

ggplot(senrep, aes(endorsee,fill=endorsee))+
  geom_bar()+
  facet_grid(~position)+
  coord_flip()+
  theme_bw()+
  theme(axis.text.x = element_text(angle = 90, hjust = 1))

More cleaning will remove undeclared endorsers and make the plot more informative.

• geom_text() adds value to each bar
• aes(label=..count..) sets value to be displayed as the count found in the dataset
• stat="count" calculates the count value called by the aesthetic
• position=position_stack(0.5) sets the position of the value text within the bar at half height

#remove undeclared endorsers, dplyr
senreps<-senrep%>%
  filter(endorsee!="undeclared")

#plot, adding value labels to each bar with geom_text and vjust
ggplot(senreps, aes(endorsee,fill=endorsee))+
  geom_bar()+
  facet_grid(~position)+
  coord_flip()+
  theme_bw()+
  geom_text(aes(label=..count..),stat="count",position=position_stack(0.5))

Biden and Booker lead in endorsements from Representatives and Senators, followed by Harris.

Further cleaning

Further cleaning will create a dataset that examines endorsements per month. dplyr can be used to aggregate by group; first, we’ll create month variable using lubridate from tidyverse, then use group_by and summarise to calculate frequencies of each endorsement.

#lubridate is used to handle dates
library(lubridate)

#three variables are created: year, month and day
endorse<-endorse%>%
  mutate(date = ymd(date))%>%
  mutate_at(vars(date),
            funs(year, month, day))


#filter out "undeclared" and endorsements from years other than 2019, from dplyr
monthly<-endorse%>%
  filter(endorsee!="undeclared" & year == 2019)

With data properly formatted, ggplot can plot a histogram.

Histogram: geom_hist

• geom_histogram() specifies type of plot
• alpha= specifies transparency on a scale from 0 to 1
• aes(y=..count..,fill=..count..) assigns color to count values with shades of blue as default
• scale_fill_gradient() specifies low and high color identities
• labs() customizes labels
• theme_classic() eliminates grid and background color
• legend.position='none' hides the legend, which proved superfluous for this plot

ggplot(monthly,aes(x=date))+
  geom_histogram(aes(y=..count..,fill=..count..,alpha=.2))+
  scale_fill_gradient("Count", low="blue", high="red")+
  labs(title="Democrat Candidate Endorsements by Date, January - April")+
  theme_classic()+
  theme(legend.position="none")

The frequency of endorsements have peaked as the candidates announce their intention to run.

Dot plot: geom_dotplot

Dot plots offer an alternative to histograms and bar plots; given their versatility they can expose hidden relationships in data.

• geom_dotplot specifies the plot
• method = "histodot" binwidth specifies a fixed bin width; for “dotdensity”, binwidth sets a maximum
• binwidth = specifies a fixed bin width with method “histodot”
• stackdir =specifies stack direction; other values are “up”, “down”, or “centerwhole”
• stackratio = specifies distance between dots in stacks, scaled from 0 to 1.
• dotsize = specifies dot size

ggplot(monthly,aes(x=date))+
  geom_dotplot(method="histodot",
               binwidth=3,
               stackdir="center",
               stackratio = .9,
               aes(color="none",
                   fill=..count..,
                   alpha=.5),
               dotsize = .90)+
  scale_fill_gradient("Count", low="green", high="red")+
  labs(title="Democrat Candidate Endorsements by Date, January - April")+
  theme_classic()+
  theme(legend.position="none")+
  scale_y_continuous(NULL,breaks=NULL)

Most endorsements were made in February; the burst of endorsements in that month coincide with Booker’s announcement, which seems to contrast with his presence in news cycles.

Another plot can be derived from the same data by defining groups as “position”; in this plot each group of endorsers are plotted separately. As a special note, the y-axis is unable to reverse with dates; May is at the top and January at the bottom.

• aex(x=factor(position), y = as.Date(date)) redundant conversion ensures date formatting
  
• origin = specifies the origin for date variables
• binaxis = "y" rotates the bin axis
• fill = is set to “position” to color based on position
• text = element_text(size=10) sets axis tick-mark font size
• xlab = or ylab = resets label text

#ensure conversion to date
monthly$date<-ymd(monthly$date)

ggplot(monthly,aes(x=factor(position), y = as.Date(date)))+
  geom_dotplot(origin=as.Date('2019-01-01'),
               method="histodot",
               binwidth=3,
               binaxis = "y",
               stackdir="center",
               stackratio = .9,
               aes(color="none",
                   fill=position,
                   alpha=.5),
               dotsize = 1)+
  labs(title="Democrat Candidate Endorsements by Date, January - April")+
  theme_classic()+
  theme(legend.position="none")+
  theme(text = element_text(size=10),
        axis.text.x = element_text(angle = 45, hjust = 1))+
  xlab("position")+
  ylab("date")

From this plot it appears that most of the endorsements have come from DNC members and members of Congress; mayors seem anxious to declare their endorsements more frequently than Senators, which is logical.

Yet another view, this time with candidates on the x-axis.

• x=factor(endorsee) groups by candidate
• fill=endorsee also groups by candidate

ggplot(monthly,aes(x=factor(endorsee), y = as.Date(date)))+
  geom_dotplot(origin=as.Date('2019-01-01'),
               method="histodot",
               binwidth=3,
               binaxis = "y",
               stackdir="center",
               stackratio = .5,
               aes(color="none",
                   fill=endorsee,
                   alpha=.1),
               dotsize = 2)+
  labs(title="Democrat Candidate Endorsements by Date, January - April")+
  theme_classic()+
  theme(legend.position="none")+
  theme(text = element_text(size=10),
        axis.text.x = element_text(angle = 45, hjust = 1))+
  xlab("position")+
  ylab("date")

One interesting fact exposed in this plot is the endorsement Biden received long before he entered the race. Additionally, endorsements for Klobuchar, Sanders, Booker and Biden were reported simultaneously by several individuals, while endorsements for Harris seem to have trickled in over the span of two months.