Linear Regression (DP IB Applications & Interpretation (AI)): Revision Note

Linear Regression

What is linear regression?

• If strong linear correlation exists on a scatter diagram then the data can be modelled by a linear model

  • Drawing lines of best fit by eye is not the best method as it can be difficult to judge the best position for the line

• The least squares regression line is the line of best fit that minimises the sum of the squares of the gap between the line and each data value

  • This is usually called the regression line of y on x

  • It can be calculated by looking at the vertical distances between the line and the data values

• The regression line of y on x is written in the form 

• a is the gradient of the line

  • It represents the change in y for each individual unit change in x

    • If a is positive this means y increases by a for a unit increase in x

    • If a is negative this means y decreases by |a| for a unit increase in x

• b is the y – intercept

  • It shows the value of y when x is zero

• You are expected to use your GDC to find the equation of the regression line

  • Enter the bivariate data and choose the model “ax + b”

  • Remember the mean point will lie on the regression line

How do I use a regression line?

• The equation of the regression line can be used to decide what type of correlation there is if there is no scatter diagram

  • If a is positive then the data set has positive correlation

  • If a is negative then the data set has negative correlation

• The equation of the regression line can also be used to predict the value of a dependent variable (y) from an independent variable (x)

  • The equation should only be used to make predictions for y

    • Using a y on x line to predict x is not always reliable

  • Making a prediction within the range of the given data is called interpolation

    • This is usually reliable

    • The stronger the correlation the more reliable the prediction

  • Making a prediction outside of the range of the given data is called extrapolation

    • This is much less reliable

  • The prediction will be more reliable if the number of data values in the original sample set is bigger