Data100 Final Prep

What is the target variable?

• The thing we're trying to predict

• The values of the y-variable in the test set.

• The x-variables

• 🎯

What's the difference between a validation set and the test set.

• The validation set is used to train the model.

• Only the validation set is used to compare models.

• The validation set is used many times, test set is used once.

Which of the sets do you not use to compare RMSE across models?

• Training

• Validation

• Testing

Which is not an assumption that we make with the train-val-test approach?

• Our sample "looks like" the population

• We are likely going to collect more data at some point

• The model we fit to the training set is a perfect representation of the population.

We use the validation set to:

• Find the best model parameters

• Find the model parameters with the best out-of-sample predictions

• Find the modelling approach that leads to the best predictions

What is information leakage?

• When any information from the test set is used in training.

• When confidential information is leaked to the public.

• A reason to call an information plumber.

• When there is some information that was lost.

We suspect that there might be an interaction between two continuous features x1 and x2. What type of plot should we make?

• A scatterplot of y versus x1, coloured by the levels of x2

• A scatterplot of y versus x1 * x2

• A scatterplot of x1 versus x2

• A boxplot of y versus x1 and x2

In a regression with body mass (g) as the target variable, bill length (nn_ has a slope of 87. What is the interpretation of this value?

• A one mm increase in bill length means 87 extra grams.

• A one g increase in body mass means 87 extra mm.

If we added bill depth (mm) to the previous regression, what would happen to the slope of body mass against bill length?

• Increase

• Decrease

• Stay the same

• ???

When making plots to explore our results, which data should we use?

• The full data so we can see exactly what's going on

• The training data so we can avoid information leakage

• The training and validation data