How to make predictions

• Anyone can make predictions.
  • Data scientists use data to inform their predictions by using the information learned from the sample to make predictions for the whole population.
• In this lab, we’ll learn how to make predictions by finding the line of best fit.
  • You will also learn how to use the information from one variable to make predictions about another variable.

Predicting heights

• Use the data() function to load the arm_span data.
• This data comes from a sample of 90 people in the Los Angeles area.
  • The measurements of height and armspan are in inches.
  • A person’s armspan is the maximum distance between their fingertips when they spread their arms out wide.
• Make a plot of the height variable.
  • If you had to predict the height of someone in the Los Angeles area, what single height would you choose and why?
  • Would you describe this as a good guess? What might you try to improve your predictions?

Predicting heights knowing arm spans

• Create two subsets of our arm_span data:
  • One for armspan >= 61 and armspan <= 63.
  • A second for armspan >= 64 and armspan <= 66.
• Create a histogram for the height of people in each subset.
• Answer the following based on the data:
  • What height would you predict if you knew a person had an armspan around 62 inches?
  • What height would you predict if you knew a person had an armspan around 65 inches?
  • Does knowing someone’s armspan help you predict their height? Why or why not?

Fitting lines

• Notice that there is a trend that people with a larger armspan also tend to have a larger mean height.
  • One way of describing this sort of trend is with a line.
• Data scientists often fit lines to their data to make predictions.
  • What we mean by fit is to come up with a line that’s close to as many of the data points as possible.
• Create a scatterplot for height and armspan. Then run the following code.

add_line()

• On the Plot pane, click two data points to draw a line through.

• NOTE: Watch the following video if you are experiencing difficulties obtaining your line:

• If you are unsuccessful using the add_line() function, refer to the next slide to learn how to use the get_line() function.

get_line()

• The get_line() function does not rely on clicking on the scatterplot to choose points, but rather on you providing the points manually.

• For example, let’s say you want to obtain the equation of the line that passes through the points (59,60) and (68,67). This is how you would use the get_line function:

get_line(c(59,60), c(68,67))

##  intercept      slope 
## 14.1111111  0.7777778

• Notice the output is the y-intercept and slope of your line.
• Now you can use the add_line() function to include the line in your scatterplot.

add_line(intercept = 14.1111111, slope = 0.7777778)

• If your line doesn’t quite fit the way you want it, try another ordered pair or make modifications to the existing equation.

Predicting with lines

• Draw a line that you think is a good fit and write down its equation. Using this equation:
  • Predict how tall a person with a 62-inch armspan and a person with a 65-inch armspan would be.

• Using a line to make predictions also lets us make predictions for armspans that aren’t in our data.
  • How tall would you predict a person with a 63.5-inch armspan to be?
• Compare your answers with a neighbor. Did both of you come up with the same equation for a line? If not, can you tell which line fits the data best?