In doomlab/learnSTATS: Introduction to Statistics (ANLY 500)

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

Programming Suggestion

• We will be using ggplot2 to create our data visualizations.
• The code can get very long, so you should use code stacking!
• Stacking your code can help you troubleshoot and see what is going on easier.

textline +
  stat_summary(fun = mean,
               geom = "point") +
  stat_summary(fun = mean,
               geom = "line", 
               aes(group = Group)) +
  stat_summary(fun.data = mean_cl_normal,
               geom = "errorbar",
               width = .2) + 
  xlab("Measurement Time") +
  ylab("Mean Grammar Score") +
  cleanup +
  scale_color_manual(name = "Texting Option",
                     labels = c("All the texts", "None of the texts"),
                     values = c("Black", "Grey")) +
  scale_x_discrete(labels = c("Baseline", "Six Months"))

Outline

• Reminder how to import files
• Two key issues for data format: factors, format

• Graphs:

  • Histograms
  • Scatterplots
  • Bar Graphs
  • Line Graphs

Working with Files

• Let's load SPSS file "ChickFlick.sav".
• Note that I have this saved in a data folder that is in the same folder as my Markdown file.
• Remember that the rio library can usually interpret these files without much work using the import() function.

library(rio)
chickflick <- import("data/ChickFlick.sav")
str(chickflick)

Factor Categorical Variables

• We saw on the previous slide that the gender and film variables have embedded data from SPSS.
• However, this does not translate to a factor variable, which is useful for making plots.

table(chickflick$gender)

table(chickflick$film)

How to Factor

chickflick$gender <- factor(chickflick$gender, #the variable you want to factor
                            levels = c(1,2), #the information already in the data
                            labels = c("Male", "Female")) #the labels for those levels

table(chickflick$gender)

Data Structure Format

• Wide datasets: rows are participants, and columns are variables.

  • Each participant gets one row, while each variable gets one column.

• Long datasets: rows are assessments, and columns are variables.

• Each time point or repeated assessment gets a separate row, while each column still represents a variable.

Rearrange Data: Wide to Long

• I wanted to test the Disney philosophy that 'Wishing upon a star makes all your dreams come true'.
• Measured the success of 250 people using a composite measure involving their salary, quality of life and how close their life matches their aspirations.
• Success was measured using a standardized technique ranging from 0 Complete failure to 4 = Complete success
• Participants were randomly allocated to either: Wish upon a star or work as hard as they can for next 5 years.
• Success was measured again after 5 years.

library(reshape) #note: you could also use pivot_longer in tidyverse
cricket <- import("data/Jiminy_Cricket.csv")
head(cricket)

Rearrange Data: Wide to Long

• id = c("column", "column") - constant variables you do not want to change. These will stay their own column but get repeated when necessary.
• measured = c("column", "column") - dependent variables you want to combine into one column.

longcricket <- melt(cricket,  #name of dataset
                    id = c("ID", "Strategy"), 
                    measured = c("Success_Pre", "Success_Post"))
#you can actually leave measured blank
head(longcricket)

Rearrange Data: Wide to Long

• Unfortunately, our melted columns become variable and value, which are not very useful names.
• We should rename them for clarity!

colnames(longcricket)[3:4] #just to figure out which ones
colnames(longcricket)[3:4] <- c("Time", "Score")

The Art of Presenting Data

• Graphs should (Tufte, 2001):

  • Show the data.
  • Induce the reader to think about the data being presented (rather than some other aspect of the graph).
  • Avoid distorting the data.
  • Present many numbers with minimum ink.
  • Make large data sets (assuming you have one) coherent.
  • Encourage the reader to compare different pieces of data.
  • Reveal data.

Why is this Graph Bad?

knitr::include_graphics("pictures/graphs/badgraph.png")

Other Bad Design Choices

• 3D charts need to be well-made (check out plotly).
• Patterns (depending)
• Cylindrical bars
• Bad axis labels
• Overlays

Why is this Graph Better?

knitr::include_graphics("pictures/graphs/bettergraph.png")

Do Not Deceive the Reader!

knitr::include_graphics("pictures/graphs/deception.png")

Plotting in R

• The great thing about ggplot2 is that it is very flexible and well documented!
• The bad thing about ggplot2 is that there's a lot going on.
• You may also need Hmisc for your error bars.

library(ggplot2)

Working with ggplot2

• How ggplot works: First you define the basic structure of a plot you want.

#an example 
myGraph <- ggplot(dataset,
                  aes(x_axis, y_axis, 
                      color = legend_var, 
                      fill = legend_var))

Working with ggplot2

• Now we have myGraph saved with all the variables, but no visualization.
• Next, you can add options to create new layers on the graph.

#an example part 2
myGraph + 
  geom_bar() +
  geom_point() +
  xlab("X Axis Label") + 
  ylab("Y Axis Label")

Histograms

• Histograms plot:

  • The continuous score (x-axis)
  • The frequency (y-axis)

• Histograms help us to identify:

  • The shape of the distribution. Measured using Skew, Kurtosis, or the Spread or variation in scores.

• Unusual scores

Histogram: Example

crickethist <- ggplot(data = cricket, #dataset
                      aes(x = Success_Pre) #only define X axis 
                      )
crickethist

Histogram: Example

• Our current plot is blank, let's add the histogram bars:

crickethist + 
  geom_histogram()

Histogram: Example

• Make the bins different:

crickethist + 
  geom_histogram(binwidth = 1)

Histogram: Example

• You can change the color:

  • Color is the outline of the bars
  • Fill is the interior of the bars

crickethist + 
  geom_histogram(binwidth = 1, color = 'purple', fill = 'magenta')

Histogram: Example

• Let's add some labels:

crickethist + 
  geom_histogram(binwidth = 1, color = 'purple', fill = 'magenta') + 
  xlab("Success Pre Test") + 
  ylab("Frequency")

Histogram: Example 2

• This dataset includes festival attendees who indicated their hygiene score on each day of the festival from 0 (eau de toilet) to 4 (eau de toilette)

festival <- import("data/festival.csv")
str(festival)

Histogram: Example 2

• In the previous example, we added things one at a time just as an example

• You can run the whole graph at once!

• Create the plot object:

festivalhist <- ggplot(data = festival, aes(x = day1)) 
festivalhist + 
  geom_histogram(binwidth = 1, color = 'blue') + 
  xlab("Day 1 of Festival Hygiene") +
  ylab("Frequency") +
  theme_bw() #theme_classic() also good!

Focus on these Facets

• All graphs should have:

  • X and Y axis labels.
  • The use of titles depends on where the graph is used.
  • Labels for the legend, facets, other group markers.
  • Error bars when appropriate.
  • Readable/cleaned up!

Clean Up?

• theme_bw() and theme_classic() are great, easy themes to make graphs presentable.
• Another important facet may be the font size, as it appears to be pretty small on the default.
• Additionally, we will want to eliminate that default gray background that is included with legends.

cleanup <- theme(panel.grid.major = element_blank(), #no grid lines
                panel.grid.minor = element_blank(), #no grid lines
                panel.background = element_blank(), #no background
                axis.line.x = element_line(color = 'black'), #black x axis line
                axis.line.y = element_line(color = 'black'), #black y axis line
                legend.key = element_rect(fill = 'white'), #no legend background
                text = element_text(size = 15)) #bigger text size

Clean Up?

• Save that code and then you can just do graph + cleanup.

festivalhist + 
  geom_histogram(binwidth = 1, color = 'blue') + 
  xlab("Day 1 of Festival Hygiene") +
  ylab("Frequency") +
  cleanup

Scatterplots

• Simple scatter: X and Y are continuous variables
• Grouped scatter: X and Y are continuous variables with legend for a third variable (generally categorical).

Scatterplots: Example

• Anxiety and Exam Performance: 103 students who rated their exam anxiety, time spent revising their exam, exam performance, and gender.

exam <- import("data/Exam_Anxiety.csv")
str(exam)

Scatterplots: Example

• Gender is currently a continuous variable, so we should factor that variable to appropriately label it on a graph when used.

table(exam$Gender)
exam$Gender <- factor(exam$Gender,
                     levels = c(1,2),
                     labels = c("Male", "Female"))
table(exam$Gender)

Simple Scatterplot

scatter <- ggplot(exam, aes(Anxiety, Exam))
scatter +
  geom_point() +
  xlab("Anxiety Score") +
  ylab("Exam Score") +
  cleanup

Simple Scatterplot with Regression Line

scatter + geom_point()+
  geom_smooth(method = 'lm', color = 'black', fill = 'blue') +
  xlab('Anxiety Score')+
  ylab('Exam Score')+
  cleanup

Grouped Scatterplot

• How to control the colors and fill with legends.

  • scale_fill_manual(name, labels, values)
  • scale_color_manual(name, labels, values)

Grouped Scatterplot with Regression Line

scatter2 <- ggplot(exam, aes(Anxiety, Exam, 
                             color = Gender, fill = Gender)) #why both?
scatter2 +
  geom_point() +
  geom_smooth(method = "lm") +
  xlab("Anxiety Score") +
  ylab("Exam Score") +
  cleanup + 
  scale_fill_manual(name = "Gender of Participant",
                    labels = c("Men", "Women"),
                    values = c("purple", "grey")) +
  scale_color_manual(name = "Gender of Participant",
                     labels = c("Men", "Women"),
                     values = c("purple", "grey10"))

GGally for Multiple Visualization

• GGally has a plotting function that uses ggplot2 as a back end.
• The function ggpairs allows you to create a scatterplot matrix for any numerical variables in a dataset.

library(GGally)
ggpairs(data = exam[ , -1], #no participant variable
        title = "Exam Anxiety, Scores, and Gender")

Bar Graphs

• Bar graphs are often used to display mean scores to allow comparison between groups.

• Error bars can be added to display the precision of the mean:

  • The confidence interval
  • The standard deviation
  • The standard error

Bar Graph: One Independent Variable

• Is there such a thing as a 'chick flick'?

  • Twenty men and twenty women were assigned to watch one of two movies: A 'chick flick' (Bridget Jones's Diary) or the control movie (Memento).
  • Physiological arousal was used as an indicator of how much they enjoyed the film.

Bar Chart: One Independent Variable

• Be sure to convert all factor variables, otherwise they will be treated as continuous (or you'll get strange errors!).

str(chickflick) #already fixed gender
chickflick$film <- factor(chickflick$film,
                    levels = c(1,2),
                    labels = c("Bridget Jones", "Memento"))

Bar Chart: One Independent Variable Example

• To add the mean, displayed as bars, we can add this as a layer to bar using the stat_summary() function:

chickbar <- ggplot(chickflick, aes(film, arousal))
chickbar + 
  stat_summary(fun = mean,
               geom = "bar",
               fill = "White", 
               color = "Black") +
  cleanup

Bar Chart: One Independent Variable Example

• To add error bars, add these as a layer using stat_summary():

chickbar + 
  stat_summary(fun = mean,
               geom = "bar",
               fill = "White", 
               color = "Black") +
  stat_summary(fun.data = mean_cl_normal, 
               geom = "errorbar", 
               position = position_dodge(width = 0.90), 
               width = 0.2) +
  cleanup

Bar Chart: One Independent Variable Example

• Now, add the rest of things we've been doing:

chickbar + 
  stat_summary(fun = mean,
               geom = "bar",
               fill = "White", 
               color = "Black") +
  stat_summary(fun.data = mean_cl_normal, 
               geom = "errorbar", 
               position = position_dodge(width = 0.90), 
               width = 0.2) +
  xlab("Movie Watched by Participant") +
  ylab("Arousal Level") +
  cleanup +
  scale_x_discrete(labels = c("Girl Film", "Guy Film"))

Bar Chart: Two Independent Variables

chickbar2 <- ggplot(chickflick, aes(film, arousal, fill = gender))
chickbar2 +
  stat_summary(fun = mean,
               geom = "bar",
               position = "dodge") +
  stat_summary(fun.data = mean_cl_normal,
               geom = "errorbar", 
               position = position_dodge(width = 0.90),
               width = .2) +
  xlab("Film Watched") +
  ylab("Arousal Level") + 
  cleanup +
  scale_fill_manual(name = "Gender of Participant", 
                    labels = c("Boys", "Girls"),
                    values = c("Gray30", "Gray"))

Line Graphs

• When to use a line graph:

  • With data that X is categorical, but is considered "mildly continuous"
  • Usually with repeated measures data over time
  • The example here is not quite appropriate for a line graph but does include repeated measures data for practicing restructuring data.

Line Graphs: One Independent Variable

• How to cure hiccups? Participants included 15 participants who tried four hiccup cures.
• The dependent variable included the number of hiccups in the minute after each procedure.

hiccups <- import("data/Hiccups.csv")
str(hiccups)

Line Graphs: One Independent Variable

• These data are in the wrong format for ggplot2 to use.
• We need all of the scores stacked up in a single column and then another variable that specifies the type of intervention.

longhiccups <- melt(hiccups, 
                    measured = c("Baseline", "Tongue", "Carotid", "Other"))
str(longhiccups)
colnames(longhiccups) <- c("Intervention", "Hiccups")

Line Graphs: One Independent Variable

hiccupline <- ggplot(longhiccups, aes(Intervention, Hiccups))
hiccupline +
  stat_summary(fun = mean, ##adds the points
               geom = "point") +
  stat_summary(fun = mean, ##adds the line
               geom = "line",
               aes(group=1)) + ##necessary for mapping line to dots
  stat_summary(fun.data = mean_cl_normal, ##adds the error bars
               geom = "errorbar", 
               width = .2) +
  xlab("Intervention Type") +
  ylab("Number of Hiccups") + 
  cleanup

Line Graphs: Two Independent Variables

• Is text-messaging bad for your grammar? Participants included 50 children who were split into two groups (between-subjects): text-messaging allowed or text-messaging forbidden.
• In a second variable (repeated measures), each child was measured at two points in time: at baseline and six months later.
• The dependent variable was the percent correct score on a grammar test.

Line Graphs: Two Independent Variables

texting <- import("data/Texting.xlsx")
str(texting)

Line Graphs: Two Independent Variables

• Two issues to clean up:

  • Correctly label the groups into a factor variable.
  • Restructure the data so the repeated measures variable is in long format.
  • When you restructure the data, you should also rename the columns

texting$Group <- factor(texting$Group,
                       levels = c(1,2),
                       labels = c("Texting Allowed", "No Texting Allowed"))
longtexting <- melt(texting,
                   id = c("Group"),
                   measured = c("Baseline", "Six_months"))
str(longtexting)
colnames(longtexting) <- c("Group", "Time", "Grammar_Score")

Line Graphs: Two Independent Variables

textline <- ggplot(longtexting, aes(Time, Grammar_Score, color = Group))
textline +
  stat_summary(fun = mean,
               geom = "point") +
  stat_summary(fun = mean,
               geom = "line", 
               aes(group = Group)) + #Group is the variable name
  stat_summary(fun.data = mean_cl_normal,
               geom = "errorbar",
               width = .2) + 
  xlab("Measurement Time") +
  ylab("Mean Grammar Score") +
  cleanup +
  scale_color_manual(name = "Texting Option",
                     labels = c("All the texts", "None of the texts"),
                     values = c("Black", "Grey")) +
  scale_x_discrete(labels = c("Baseline", "Six Months"))

Summary

• In this lecture, you have learned:

  • How to clean up your data for visualizations: factor() and melt()
  • Ideas for good visualizations
  • Histograms, scatterplots, bar charts, and line graphs
  • ggplot2 structure and how to get started

doomlab/learnSTATS documentation built on June 9, 2022, 12:54 a.m.