Load libraries.

library(stringr)
library(tidyr)
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, union

library(ggplot2)

Read in a CSV file with survey data about Thanksgiving.

More info available here:

https://github.com/fivethirtyeight/data/tree/master/thanksgiving-2015

csv_link <- "https://raw.githubusercontent.com/fivethirtyeight/data/master/thanksgiving-2015/thanksgiving-2015-poll-data.csv"
thanksgiving <- read.csv(csv_link,header=TRUE,stringsAsFactors=FALSE,check.names=FALSE)

Filter only for individuals that celebrate Thanksgiving, after cleaning up some column names.

colnames(thanksgiving)[2] <- "Celebrates_Thanksgiving"
colnames(thanksgiving)[c(25,38,50)] <- c("Which of these side dishes aretypically served at your Thanksgiving dinner? Please select all that apply. - Other (please specify again)",
"Which type of pie is typically served at your Thanksgiving dinner? Please select all that apply. - Other (please specify again)",
"Which of these desserts do you typically have at Thanksgiving dinner? Please select all that apply.   - Other (please specify again)")
thanksgiving <- thanksgiving %>% filter(Celebrates_Thanksgiving == "Yes")

Select columns corresponding to side dishes, minus “Other”.

Then, count number of people who selected each side.

side_dish_columns <- grep('Which of these side dishes aretypically served at your Thanksgiving dinner?',colnames(thanksgiving),value=TRUE)
side_dish_columns <- side_dish_columns[grep('Other',side_dish_columns,invert=TRUE)]

side_dish_frequency <- data.frame(Side = rep(NA,times=length(side_dish_columns)),
Frequency = rep(NA,times=length(side_dish_columns)))

for(i in 1:length(side_dish_columns))
{
side_name <- str_replace_all(side_dish_columns[i],
pattern='Which of these side dishes aretypically served at your Thanksgiving dinner\\? Please select all that apply\\. - ',
replace='')

side_dish_frequency[i,] <- c(side_name,length(which(thanksgiving[,side_dish_columns[i]] != "")))
}

side_dish_frequency$Frequency <- as.numeric(side_dish_frequency$Frequency)

Make a barplot of the frequency of different sides, minus “Other”.

ggplot(side_dish_frequency,aes(Side,Frequency)) + 
geom_bar(stat="identity") + 
theme(axis.text.x = element_text(angle = 90, hjust = 1))

Now use scale_fill_manual to color each bar by a color appropriate to the food.

Use colors here as a guide: http://www.stat.columbia.edu/~tzheng/files/Rcolor.pdf

Put in colors based on order sides appear in side_dish_frequency.

Let’s color fruit salad as red, since may have watermelon, strawberry, etc.

ggplot(side_dish_frequency,aes(Side,Frequency,fill=Side)) + 
geom_bar(stat="identity") + 
theme(axis.text.x = element_text(angle = 90, hjust = 1)) +
scale_fill_manual(values = c("darkgreen","darkorange","floralwhite","gold","gold","firebrick1","darkgreen","gold","floralwhite","floralwhite","darkorange","darkgreen","darkorange"))

Sort by frequency, then adjust the order of the color vector accordingly.

color_per_side <- c("darkgreen","darkorange","floralwhite","gold","gold","firebrick1","darkgreen","gold","floralwhite","floralwhite","darkorange","darkgreen","darkorange")
color_per_side <- color_per_side[order(side_dish_frequency$Frequency)]

side_dish_frequency$Side <- factor(side_dish_frequency$Side,
levels=side_dish_frequency$Side[order(side_dish_frequency$Frequency)])

ggplot(side_dish_frequency,
aes(Side,Frequency,fill=Side)) + 
geom_bar(stat="identity") + 
theme(axis.text.x = element_text(angle = 90, hjust = 1)) +
scale_fill_manual(values = color_per_side)

We find that the two most popular Thanksgiving sides are both white.

What if we use shape_colour_manual instead, and specify color by side instead of fill by side?

ggplot(side_dish_frequency,
aes(Side,Frequency,colour=Side)) +
geom_bar(stat="identity",fill="lightgrey") + 
theme(axis.text.x = element_text(angle = 90, hjust = 1)) +
scale_colour_manual(values = color_per_side)

Now the color of the border of the bars will change based on the colors we have specified.

Using “colour” instead of “fill” probably isn’t what we want for a barplot. But it might be useful for other types of plots (like scatterplots with a hollow dot for the points).