Regressions GMLE Practice Questions (Page 4)

Questions 16–20

1. 16application · medium

   A hospital wants to predict the probability that a patient has a certain disease based on several biomarkers. The dataset is imbalanced: only 5% of patients have the disease. The model must output a probability. Which evaluation metric is most appropriate?

   • AR-squared
   • BAccuracy
   • CMean squared error (MSE)
   • DArea under the ROC curve (AUC-ROC)
   Reveal Solution Discussion Select an answer first
2. 17application · medium

   A team is building a linear regression model to predict employee performance from 50 highly correlated features. The model has high variance on the validation set. The team wants to reduce overfitting while keeping all features in the model. Which approach should they choose?

   • AElastic net regression
   • BLasso regression
   • CPolynomial regression
   • DRidge regression
   Reveal Solution Discussion Select an answer first
4. 18foundation · easy

   What is the primary risk of using a high-degree polynomial regression model?

   • AUnderfitting the training data, resulting in a model that is too simple
   • BThe model will always have a higher R-squared on test data
   • COverfitting the training data, leading to poor generalization on unseen data
   • DThe model becomes non-linear and cannot be trained with least squares
   Reveal Solution Discussion Select an answer first
5. 19expert · hard

   A data scientist is building a logistic regression model for a binary classification problem. The model's coefficients are very large, and the training loss is low, but the validation loss is high. What is the most likely issue and the best remedy?

   • AThe learning rate is too high; reduce it.
   • BOverfitting; apply L2 regularization to shrink the coefficients.
   • CUnderfitting; increase the number of features.
   • DThe decision boundary is wrong; change the threshold.
   Reveal Solution Discussion Select an answer first
6. 20application · medium

   A team is fitting a polynomial regression model with degree 15 to a small dataset. The training R-squared is 0.99, but the validation R-squared is 0.60. Which action is most appropriate to improve generalization?

   • AUse ordinary least squares without regularization.
   • BRemove all polynomial terms and use a linear model.
   • CIncrease the polynomial degree to 20.
   • DApply lasso regression with a suitable alpha.
   Reveal Solution Discussion Select an answer first