Question

When you use linear regression to fit a linear model, and create a scatterplot of actual vs. predicted values, you would ideally see: a. the points lie close to the diagonal line from bottom left to upper right b. the points form a random "cloud" C. the point lie close to a horizontal line (write a, b or c): (True/False) If you have many variables (features), you will tend to prefer non-parametric methods to parametric methods. The two plots below show residuals vs. fitted plots from two different linear models created from the same data set. Which of the two models would you tend to prefer based on just these plots? (write a or b) a. the model of the left plot b. the model of the right plot Residuals vs Fitted Residuals vs Fitted 2000 2000 100 200 032 L U Residuals 000 300-200 -100 83 0 200 400 600 800 1000 0 200 400 600 800 1000 Fitted values Fitted values

Answer 1

When ... : (a) the points ... right.

If the model is already fiited then actual and predicted values will be almost same.

If ... : True

Non parametric models do not have a fixed shape and thus suit large number of variables.

The ... : (a)

The model on left plot has less mean squared error.