Finding patterns in data.

• To discover (really) interesting observations or relationships in data, we need to find them!
  • Which is difficult if we only look at the raw data.
• The best tool for finding patterns is often … your own eyes.
  • Plots are an excellent way to help your eye search for patterns.
• In this lab, we’ll learn how to include more variables in our plots to make them more informative.
• Import the data from your class’s Food Habits campaign and name it food.

Where are the variables?

• How many variables were used to create this plot? Which variables were used and how were they used?

Multiple variable plots

• The previous graph is an example of a multiple variable plot, which means that more than a single variable was used. In this case:
  • Variable 1: height
  • Variable 2: sex
• Multiple variable plots are tools for finding relationships between data.
• Let’s take our food data and make some new multiple variable plots you haven’t created before!

Scatterplots

• Scatterplots are useful for viewing how one numerical variable relates to another numerical variable.

Creating scatterplots

• Fill in the blanks to create a scatterplot with sodium on the y-axis and sugar on the x-axis.

xyplot(____ ~ ____, data = food)

Scatterplots in action

• Use a scatterplot to answer the following questions:
  • Do snacks that have more protein also have more calories? Why do you think that?
  • What happens if you swap the protein and calories variables in your code? Does the relationship between the variables change?
  • Does the relationship between protein and calories change when the snack is either Salty or Sweet? Write down the code you used to answer this question.

4-variable scatterplots

• When we make scatterplots, we can include:
  • 1 numerical variable on the x-axis
  • 1 numerical variable on the y-axis
  • Use 1 categorical variable to facet our scatterplot
  • Change the color of the points based on another categorical variable
• To change the color of our points, we can include the groups argument much like we did for bargraphs (use the search feature in the History pane if you need help).
• Create a scatterplot that uses these 4 variables: sodium, sugar, cost, salty_sweet.

Multiple facets

• It can sometimes be helpful to facet on more than 1 variable.
  • Splitting the data using 2 facets can give us additional insights that might otherwise be hidden.
• Create a dotPlot or histogram of the calories variable, but facet the data using:

healthy_level + salty_sweet 

• How does the healthy_level of a Salty or Sweet snack impact the number of calories in the snack?

• Although we are treating healthy_level as a categorical variable, R recognizes it as a numerical variable.

  • Verify this using the str function.
  • Notice that the faceted histograms or dotPlots do not have labels, but rather tick-marks.
  • You will have the opportunity to convert the healthy_level variable into a factor later on.

• Faceting your data on a numerical variable is NOT recommended.

  • Numerical variables often have so many different values that they overwhelm the plot and make it hard to read.

On your own

• Answer the following questions by creating an appropriate graph or graphs:
  • Do healthier snacks have more or less ingredients than less healthy snacks?
  • What other variables seem to be related to the number of ingredients of a snack? Describe their relationships.