R Workshop: Data Visualization and Management

Author

Brier Gallihugh, M.S.

Published

June 19, 2023

Installing The `Tidyverse` Package

install.packages("tidyverse")
library(tidyverse)

The above code is how you want to start any R script. You always want to install and load in any packages that you may need in order to run analyses. For this part of the R workshop we will be working with what is called the tidyverse package. It’s essentially the go to array of packages in R for data science needs (and therefore a good portion of our needs as well)

A Note On Packages & library() Function

You only need to install a package once (unless you update your version of R). The package gets stored on your local computer. A library() function call imports the installed package from your local storage. Further, you only need to call the library() function once per R script

Working With The `dplyr` Package (Data Manipulation)

library(tidyverse)
library(skimr)

dplyr_data <- dplyr::starwars

1: Call the tidyverse packages
2: We will be using the starwars data set for the dplyr tutorial. I’ve assigned it to the variable dplyr_data here.

skim(dplyr_data)

1: We can view some of the key variable data using the skim()

Data summary
Name	dplyr_data
Number of rows	87
Number of columns	14
_______________________
Column type frequency:
character	8
list	3
numeric	3
________________________
Group variables	None

Variable type: character

skim_variable	n_missing	complete_rate	min	max	n_unique
name	0	1.00	3	21	87
hair_color	5	0.94	4	13	12
skin_color	0	1.00	3	19	31
eye_color	0	1.00	3	13	15
sex	4	0.95	4	14	4
gender	4	0.95	8	9	2
homeworld	10	0.89	4	14	48
species	4	0.95	3	14	37

Variable type: list

skim_variable	complete_rate	n_unique	min_length	max_length
films	1	24	1	7
vehicles	1	11	0	2
starships	1	17	0	5

Variable type: numeric

skim_variable	n_missing	complete_rate	mean	sd	p0	p25	p50	p75	p100	hist
height	6	0.93	174.36	34.77	66	167.0	180	191.0	264	▁▁▇▅▁
mass	28	0.68	97.31	169.46	15	55.6	79	84.5	1358	▇▁▁▁▁
birth_year	44	0.49	87.57	154.69	8	35.0	52	72.0	896	▇▁▁▁▁

head(dplyr_data)

1: We can also use the head() function which simply gives you a print out of the first 5 rows of a data set.

Recoding Variables

One variable when looking at the starwars data set might be sex. Here we can see it is coded as both a character and as either male, female or NA. For a simple recode we might wish to

Transform the variables into a factor
Change the naming convention to maybe 1, 0 and Unknown

We can achieve this with the code below

dplyr_data <- dplyr_data %>% 
  mutate(sex = as.factor(sex),
         sex = recode(sex,
                      'male' = '0',
                      'female' = '1'))

1: To recode the variable sex we need to use the mutate() function and as.factor() functions as shown above
2: To recode the values for male and female to 0 and 1 respectively, we need to use the recode() function as shown here. ::: {.callout-tip} ### A Note On %>% Operator You may have noticed this %>% operator. This is a handy operator that essentially takes the data on the left hand side and “pipe”s it into whatever is on the right as the first argument. This is most effective when the right hand function is expecting some form of a data set :::

A Note On dplyr::recode() Function

The recode() function in the dplyr package uses what is called OLD to NEW syntax. This just means that when renaming variables as shown here, you want to list the original variable name followed by you new desired variable name

Creating Variables

Creating variables in R can be done a couple of ways. One is a little clunky (from a code perspective) and the other is more elegant. I’ll cover the more clunky way first followed by the more elegant way second. I’ll illustrate this by creating a variable that takes the mass variable from the starwars data set and reduces it by 10 units

dplyr_data$mass_10a <- dplyr_data$mass - 10


dplyr_data <- dplyr_data %>%
  mutate(mass_10b = mass - 10)

1: This ways is relatively simply because you can think of it as a simple formula notation. However, it’s a little clunky because typically adding a $ operator is considered poor coding practice
2: The more elegant way to create a variable is to simply again use the mutate() function

A Note On the $ Operator

The $ operator simply says from the data set on the left of the operator, please find (or create) the variable on the right. In this case, from the dplyr_data data, create the variable mass_10a

Filtering Variables

Keeping with the starwars data set, we might wish to revisit our earlier mutate of the male and female sex variable categories. Suppose for an analysis we wish to only include the male and female starwars characters? For this we might wish to filter so that our data only contains males and females. The code below will illustrate exactly how to do this

dplyr_data_mf <- dplyr_data %>%
  filter(sex == 1 | sex == 0)

1: Here we have a filter() function that takes an argument for which conditions to include [==]. In this case we have when sex = 1 OR [|] when sex = 0.

A Note on Syntax

The filter() function uses the notation == to serve as “equals”. You may also tell filter() what NOT to include with the notation !=

Reverse Coding

It is not uncommon for many of you to work with scales that might require some form of reverse coding. This can be accomplished using the following syntax. What is left will be the original dataframe with added columns for the items that we’ve reverse coded. They will have a “_R” variable name for ease of use

library(tidyverse)
library(psych)

df <- data.frame(Q1 = c(1,3,4,5,6,7),
                 Q2 = c(3,4,5,5,7,7),
                 Q3 = c(1,2,2,4,1,1))

reverse_key <- c(1,-1,1)

df_R <- data.frame(reverse.code(keys = reverse_key,
                   items = df[,c("Q1","Q2","Q3")],
                   mini = 1,
                   maxi = 7)) %>%
  rename("Q2_R" = "Q2.")

df <- right_join(df,df_R,
                keep = FALSE)

print(df)

1: The psych package contains a reverse.code() function for scale items
2: There are no convenient pre-built data sets for this so I’ve created a quick toy one called df with the variables Q1, Q2, and Q3
3: The reverse.code() function requires a keys argument which is essentially a numerical vector of length of the reverse coded items that correspond sequentially to which items are (-1) and aren’t (1) reverse coded
4: This is the start of the reverse.code() function within a new dataframe
5: I’ve subset (only included) the scale items here using this notation
6: Mini refers to the lowest possible value for the scale (i.e., 1)
7: Maxi refers to the highest possible value for the scale (i.e., 7)
8: I’ve added a rename() function to rename the reverse coded items from “ItemX.” to “ItemX_R” so we can track which items are the reverse coded one’s later
9: Joining the two data frames into the original one so we only have to worry about the original data set
10: Refers to keeping the keys used to join the two data frames (i.e., unique identifiers). We don’t want to keep them here

  Q1 Q2 Q3 Q2_R
1  1  3  1    5
2  3  4  2    4
3  4  5  2    3
4  5  5  4    3
5  6  7  1    1
6  7  7  1    1

Working With The `stringr` Package (Working w/ Strings)

The stringr package is primarily used when working with what are known as strings of data. Essentially text box types of free response options. For example maybe in a Qualtrics form you allow someone to list “Other” as their religious belief system but ask them upon that selection choice to type out a better word. Same might be true for gender for example. Below we’ll use the words data set to some basic text manipulation with the first 10 rows of data. On the right, we will see our original data set. However, on the right we will see that data set ultimately filtered by whether or not there is an ab in the words variable for a given observation.

Text Detection With `stringr` Package

library(tidyverse)

stringr_data <- data.frame(stringr::words %>%
                             head(10)) %>%
  rename("words" = "stringr..words.....head.10.")

stringr_data_original <- stringr_data %>%
  mutate(match = str_detect(words,pattern = "ab"))


stringr_data_new <- stringr_data_original %>%
  filter(match == TRUE)

1: Here I am converting the stringr data into a data frame and selecting the first 10 observations for simplicity.
2: I’m also using the rename() function to change the preset variable name to “words”.
3: I’m “piping” the stringr_data into the mutate() function
4: This line shows that I am creating a variable called match that will output a TRUE or FALSE if in the column words there is a pattern of "ab".
5: I am filtering the column match by whether or not it is TRUE (i.e., whether an observation consists of the pattern “ab”)

Original Output

      words match
1         a FALSE
2      able  TRUE
3     about  TRUE
4  absolute  TRUE
5    accept FALSE
6   account FALSE
7   achieve FALSE
8    across FALSE
9       act FALSE
10   active FALSE

New Output

     words match
1     able  TRUE
2    about  TRUE
3 absolute  TRUE

Text Replacement With The `stringr` Package

While we’ve seen how to pull out matching observations using text responses, maybe we want to actually modify the responses. We can do that as well. We will demonstrate using the new data frame consisting of 3 words. Let’s as an example replace the pattern “ab” with nothing. We see how to do that below

stringr_data_new <- stringr_data_new %>% 
  mutate(across(.cols="words",
                .fns=str_replace,
                pattern = "ab",
                replacement = ""))

print(stringr_data_new)

1: Here I am specifing that I wish to apply a function to the words column
2: The function I wish to apply is the str_replace function which takes two arguments (pattern and replacement which I’m about to specify)
3: I specify the pattern I’m looking for as "ab"
4: I specify what I would like to replace that pattern with. In this case I don’t want anything so I just put ""

   words match
1     le  TRUE
2    out  TRUE
3 solute  TRUE

Working With The `lubridate` Package (Date Data)

Personally, I don’t work with date data very often. Usually time simply isn’t a variable I’m interested in. However, for many of you who may be clinical or health focused, this is likely not your experience. Lets see how we can use the lubridate package to mess with date formatted data

Converting to Date Format

library(tidyverse)

lubridate_data <- lubridate::lakers

lubridate_data <- lubridate_data %>% 
  mutate(across(.cols=date,
                .fns=ymd)) %>%
  mutate(date_myd = format(as.Date(date),"%m-%d-%Y"))

1: Here I am saying I wish to apply the function ymd() to the date column
2: For this line, I am saying I wish to create a new variable called date_myd by formatting the date variable both as a date AND then formatted to a mm-dd-yyyy format. That corresponds to the "%m-%d-%Y" string we see on this line.

Modifying Date Format

We can see here that we’ve converted a numeric value in the format (YYYYMMDD) into a date in the “Year-Month-Date” format. This even looks a little more appealing to the eye especially as you’re scanning the date. However, what if you don’t like YYYY-MM-DD format and would rather have something like MM-DD-YYYY format instead as is common in the US? Below you can see how to take the format we just used and convert it to the more US common syntax shown on the left. On the right, we can see how to do it for the more EU common syntax of DD-MM-YY

lubridate_data <- lubridate_data %>%
  mutate(date_dmy = format(as.Date(date),"%d-%m-%Y"))

1: Here I am doing the same as earlier but I am changing the format code to be dd-mm-yyyy using the string "%d-%m-%Y"

May-20-2023 Format

[1] "10-28-2008"

20-May-2023 Format

[1] "28-10-2008"

Working With The `ggplot2` Package

Standard Histogram With Density Curve

library(tidyverse)
library(jtools)
  
gender <- rep(c("male","female"),50)
test <- rnorm(100,mean = 75,sd=2)
  
df <- data.frame(gender,test)

density_plot <- ggplot(df,aes(x = test)) +
  geom_histogram(aes(y=after_stat(density)),binwidth = 1) +
  stat_function(fun = dnorm,
                args = list(mean = mean(test),
                            sd = sd(test)),
                col = "blue",
                linewidth = 1) +
  jtools::theme_apa() +
  labs(title = "Figure 1. Histogram of Test Scores",
       x = "Test Scores",
       y = "Score Density")

ggsave("histogram.png")
print(density_plot)

1: Creation of a basic data set consisting of 100 observations of 2 variables (gender and test)
2: Initial ggplot2 taking the arguments for df as the data and test as our variable to create a histogram of
3: The geom_histogram() tells ggplot2 what type of geom to draw using the aes() data above. The aes(y=after_stat(density)) tells ggplot to convert the y axis as a function of density (vs count which is the default)
4: This stats_function allows us to graph a statistic onto the graph. In this case we want it to graph a normal distribution (the dnorm function) of the variable we care about.
5: The stats_function takes an args() function that we have to give it the mean and sd of the variable we care about. This is shown here
6: These provide some general aesthetic choices so we’ve specified the curve to be colored blue with a relatively small line width of 1.
7: The theme_apa() function simply modifies the ggplot2 graph to roughly align with APA formatting
8: The labs() function allows us to add labels to our prospective histogram
9: This will save the built graphic as a .png file
10: This will print the ggplot2 column plot

Standard Column Bar Graph

library(tidyverse)
library(jtools)

col_data <- mtcars

skimr::skim(col_data)

1: The mtcars data set comes with the ggplot2 package. Finally I used the skim() function to take a quick look at the data

Data summary
Name	col_data
Number of rows	32
Number of columns	11
_______________________
Column type frequency:
numeric	11
________________________
Group variables	None

Variable type: numeric

skim_variable	complete_rate	mean	sd	p0	p25	p50	p75	p100	hist
mpg	1	20.09	6.03	10.40	15.43	19.20	22.80	33.90	▃▇▅▁▂
cyl	1	6.19	1.79	4.00	4.00	6.00	8.00	8.00	▆▁▃▁▇
disp	1	230.72	123.94	71.10	120.83	196.30	326.00	472.00	▇▃▃▃▂
hp	1	146.69	68.56	52.00	96.50	123.00	180.00	335.00	▇▇▆▃▁
drat	1	3.60	0.53	2.76	3.08	3.70	3.92	4.93	▇▃▇▅▁
wt	1	3.22	0.98	1.51	2.58	3.33	3.61	5.42	▃▃▇▁▂
qsec	1	17.85	1.79	14.50	16.89	17.71	18.90	22.90	▃▇▇▂▁
vs	1	0.44	0.50	0.00	0.00	0.00	1.00	1.00	▇▁▁▁▆
am	1	0.41	0.50	0.00	0.00	0.00	1.00	1.00	▇▁▁▁▆
gear	1	3.69	0.74	3.00	3.00	4.00	4.00	5.00	▇▁▆▁▂
carb	1	2.81	1.62	1.00	2.00	2.00	4.00	8.00	▇▂▅▁▁

col_data <- col_data %>%
  group_by(cyl) %>%
  summarize(n = n(),
            mpg_average = mean(mpg))

col_plot <- ggplot(col_data,aes(x = as.factor(cyl),
                                y = mpg_average,
                                fill = as.factor(cyl))) +
  geom_col(color = "black") +
  labs(x = "Number of Cylinders", 
       y = "Average Fuel Economy (mpg)",
       title = "Figure 2. Average Fuel Economy by Cylinder Count",
       caption = "Source: Data from the mtcars data set") +
  jtools::theme_apa() +
  theme(plot.caption = element_text(hjust = 0)) +
  scale_fill_manual(values = c("grey50","grey80","grey100"))

ggsave("col_plot.png")
print(col_plot)

2: I want to modify my data so that I have it grouped by cyl and n() and average_mpg are calculated
3: I’m starting to layer my column plot with this. aes() is where you put your important data (e.g., x and y variables)
4: The geom_col() tells ggplot2 what type of geom to draw using the aes() data above
5: The theme(plot.caption = element_text(hjust = 0)) just left justifies the caption
6: The scale_fill_manual() tells ggplot2 what to assign for the fill variable in the aes() function

Standard Boxplot Graph

library(tidyverse)
library(jtools)
bplot_data <- mtcars

box_plot <- ggplot(bplot_data,aes(x = as.factor(cyl),
                                  y = mpg)) + 
  geom_boxplot(outlier.shape = NA) +
  labs(x = "Number of Cylinders", 
       y = "Average Fuel Economy (mpg)", 
       title = "Figure 3. Boxplot of Distribution of Average Fuel Economy by Cylinder Count",
       caption = "Source: Data from the mtcars data set") +
  jtools::theme_apa() +
  theme(plot.caption = element_text(hjust = 0)) +
  scale_fill_manual(values = c("grey50","grey80","grey100"))

ggsave("box_plot.png")
print(box_plot)

1: The beauty of ggplot2 is that there is a lot of overlap between different geom. The data to make a column chart vs a box plot in ggplot2 is just the geom_boxplot vs geom_col function calls shown here

Standard Violin Plot

library(tidyverse)
library(jtools)
violin_data <- mtcars

violin_plot <- ggplot(violin_data,aes(x = as.factor(cyl),
                                      y = mpg,
                                      fill = as.factor(cyl))) + 
  geom_violin(draw_quantiles = c(.25,.50,.75)) +
  labs(x = "Number of Cylinders",
       y = "Average Fuel Economy (mpg)",
       title = "Figure 4. Violin Plot of Distribution of Average Fuel Economy by Cylinder Count",
       caption = "Source: Data from the mtcars data set") +
  jtools::theme_apa() +
  theme(plot.caption = element_text(hjust = 0)) + 
  scale_fill_manual(values = c("grey50","grey80","grey100"))

ggsave("violin.png")
violin_plot

1: The draw_quartiles function takes a numeric list to represent the quartiles you want. I’ve chosen the most common of 25%, 50% and 75% but you can input any set of 3 values you’d like

Standard Line Graph

library(tidyverse)
library(jtools)
library(skimr)

line_data <- txhousing
skimr::skim(line_data)

1: We’re now using a Texas housing data set found the ggplot2 package. We can take a look at it by using the skim() function in the skimr package

Data summary
Name	line_data
Number of rows	8602
Number of columns	9
_______________________
Column type frequency:
character	1
numeric	8
________________________
Group variables	None

Variable type: character

skim_variable	n_missing	complete_rate	min	max	empty	n_unique	whitespace
city	0	1	4	21	0	46	0

Variable type: numeric

skim_variable	n_missing	complete_rate	mean	sd	p0	p25	p50	p75	p100	hist
year	0	1.00	2007.30	4.50	2000	2003.00	2007.00	2011.00	2015.0	▇▆▆▆▅
month	0	1.00	6.41	3.44	1	3.00	6.00	9.00	12.0	▇▅▅▅▇
sales	568	0.93	549.56	1110.74	6	86.00	169.00	467.00	8945.0	▇▁▁▁▁
volume	568	0.93	106858620.78	244933668.97	835000	10840000.00	22986824.00	75121388.75	2568156780.0	▇▁▁▁▁
median	616	0.93	128131.44	37359.58	50000	100000.00	123800.00	150000.00	304200.0	▅▇▃▁▁
listings	1424	0.83	3216.90	5968.33	0	682.00	1283.00	2953.75	43107.0	▇▁▁▁▁
inventory	1467	0.83	7.17	4.61	0	4.90	6.20	8.15	55.9	▇▁▁▁▁
date	0	1.00	2007.75	4.50	2000	2003.83	2007.75	2011.67	2015.5	▇▇▇▇▇

line_data <- line_data %>%
  group_by(year) %>%
  summarize(total_sales = sum(sales, na.rm = TRUE))

line_plot <- ggplot(line_data,aes(x = year,
                                  y = total_sales)) +
  geom_point() +
  geom_line() +
  labs(x = "Year",
       y = "Total Housing Sales",
       title = "Figure 5. Total Texas Housing Sales By Year",
       caption = "Source: Data from the ggplot2 data set") +
  scale_x_continuous(breaks = seq(2000,2015,2)) +
  jtools::theme_apa() +
  theme(plot.caption = element_text(hjust = 0))

ggsave("line.png")
print(line_plot)

2: It might be useful to see how sales have changed over time within Texas. As such we might want to summarize the total number of home sales by year. How to do this is illustrated here with a group_by() and summarize() function.
3: We need to feed the ggplot object our aes() variables. For this we’ve selected year and total_sales as our x and y variable respectively
4: We might want to add points to our line graph for readability so we can add a geom_point() layer
5: Now we want to add our actual lines. We can do that by providing a geom_line() layer
6: Again we are adding our typical labels here
7: This scale_x_continous variable might seem weird. However if we look at our data we will see that our year variable is continuous rather than categorical. Further, the initial breaks skip by intervals of 5 between 2000 and 2015. As such, we may want to change this. We can do that with this function call. The seq function allows us to dictate the min and max of the x values and how we scale our graph. I’ve choosen to go by increments of 2.

Standard Column Bar Graph W/ Std. Error

library(tidyverse)
library(jtools)
col_SE_data <- mtcars

col_SE_data <- col_SE_data %>%
  group_by(cyl) %>%
  summarize(n = n(),
            mpg_average = mean(mpg, na.rm = TRUE),
            sd = sd(mpg, na.rm = FALSE),
            se = sd/sqrt(n))

col_SE_plot <- ggplot(col_SE_data,aes(x = as.factor(cyl),
                                      y = mpg_average,
                                      fill = as.factor(cyl))) +
  geom_col(color = "black") +
  geom_errorbar(aes(ymax = mpg_average + se,
                    ymin = mpg_average - se), width = .5) +
  labs(x = "Number of Cylinders",
       y = "Average Fuel Economy (mpg)",
       title = "Figure 6. Average Fuel Economy by Cylinder Count",
       caption = "Source: Data from the mtcars data set") +
  scale_fill_manual(values = c("grey50","grey80","grey100")) +
  jtools::theme_apa() + 
  theme(plot.caption = element_text(hjust = 0))

ggsave("col_se.png")
print(col_SE_plot)

1: For the standard error chart, we have to borrow a bit from our previous line chart syntax as we need to manually compute some group level statistics in order to calculate SE. Here we’re grouping by cyl and we need to compute the n and SD to compute the SE. This syntax shows how to do this
2: We need to provide our aes() factors. Here we want cyl,mpg_average and a fill aesthetic (for color)
3: We need to add our standard geom_col layer
4: For our error bars, we want to call geom_errorbar and designate our ymax (upper level) and ymin (lower level) bands. This will do that
5: Adding our usual labels
6: Modify our colors for the column

Installing The Tidyverse Package

Working With The dplyr Package (Data Manipulation)

Recoding Variables

Creating Variables

Filtering Variables

Reverse Coding

Working With The stringr Package (Working w/ Strings)

Text Detection With stringr Package

Original Output

New Output

Text Replacement With The stringr Package

Working With The lubridate Package (Date Data)

Converting to Date Format

Modifying Date Format

May-20-2023 Format

20-May-2023 Format

Working With The ggplot2 Package

Standard Histogram With Density Curve

Standard Column Bar Graph

Standard Boxplot Graph

Standard Violin Plot

Standard Line Graph

Standard Column Bar Graph W/ Std. Error

Installing The `Tidyverse` Package

Working With The `dplyr` Package (Data Manipulation)

Working With The `stringr` Package (Working w/ Strings)

Text Detection With `stringr` Package

Text Replacement With The `stringr` Package

Working With The `lubridate` Package (Date Data)

Working With The `ggplot2` Package