Unit 9: Regression Models - Practice Quiz

1 What is the main purpose of regression analysis?

2 In regression, what is the variable being predicted called?

3 Which task is most suitable for a regression model?

4 What does simple linear regression use to predict an outcome?

5 Which regression type uses two or more predictor variables?

6 Which type of regression can represent a curved relationship by including powers of a predictor?

7 Which equation represents a simple linear regression model?

8 In the equation , what does represent?

9 What is a residual in linear regression?

10 What does fitting a linear regression model usually try to minimize?

11 Logistic regression is mainly used for which kind of task?

12 What range of values does the sigmoid function produce?

13 In binary logistic regression, how many possible target classes are there?

14 What is the main purpose of regularization?

15 Which regularization method uses the absolute values of model coefficients?

16 Which regularization method uses squared model coefficients in its penalty?

17 What does Mean Absolute Error (MAE) measure?

18 Which metric gives larger errors more weight by squaring them?

19 What does an value close to generally indicate?

20 Which metric is commonly used to measure the proportion of correct logistic regression classifications?

21 A regression model predicts a house price of , while the actual price is . If the residual is defined as actual minus predicted, what is the residual?

22 A residual plot shows residuals forming a clear curved pattern around zero. What does this most strongly suggest?

23 A researcher wants to predict salary using years of experience, education level, and number of certifications. Which regression type is most appropriate if the response is continuous?

24 The relationship between temperature and electricity demand is U-shaped. Which model can represent this pattern directly?

25 Why is polynomial regression considered a form of linear regression?

26 A fitted model is . What prediction does it produce when ?

27 In the model , how should the coefficient of be interpreted?

28 Two predictors have ranges to and to . Why might standardizing them help when fitting linear regression with gradient descent?

29 A linear model performs well on training data but poorly on validation data after many irrelevant predictors are added. What is the most likely issue?

30 A logistic regression model outputs for a customer. What does this value normally represent?

31 Using a classification threshold of , which class is assigned to an observation with predicted probability ?

32 In logistic regression, a predictor has coefficient . Holding other variables constant, how does a one-unit increase in that predictor affect the odds?

33 A fraud detector misses too many fraudulent transactions. Which threshold change is most likely to increase recall for the fraud class?

34 Which regularization method is most suitable when the goal is to shrink coefficients while generally retaining all predictors?

35 A dataset contains many predictors, but only a small subset is believed to be useful. Which method can perform implicit feature selection?

36 As the regularization parameter becomes very large, what usually happens to a regularized regression model?

37 A regression model produces errors , , and . What is its mean absolute error?

38 A model has on a test set. Which interpretation is most appropriate?

39 A classifier has true positives and false positives. What is its precision?

40 Two regression models have similar MAE values, but Model A has a much larger RMSE. What is the most likely explanation?

41 In the population model , suppose . A researcher omits and regresses only on . When is the resulting slope for unbiased for ?

42 A regression model has excellent out-of-sample predictive accuracy, but the training data are observational and contain an unmeasured confounder affecting both and . Which conclusion is justified?

43 A response is a nonnegative count with conditional variance substantially larger than its conditional mean, even after relevant predictors are included. Which model is most directly appropriate?

44 Which regression method estimates the conditional median of given and is comparatively resistant to extreme response values?

45 A model is written as . How should it be classified?

46 In an ordinary least-squares model, one predictor is an exact linear combination of two others. Which statement is correct?

47 Assume , but varies with . Under otherwise standard conditions, what is true of ordinary least squares?

48 A linear model includes , , and their interaction . After replacing by while retaining all corresponding terms, what remains unchanged?

49 A model for has residuals . If the fitted log-response is , what estimates the conditional mean under this assumption?

50 In a logistic model , what does represent?

51 A predictor perfectly separates the two outcome classes in unregularized logistic regression. What typically happens to its maximum-likelihood coefficient estimate?

52 A calibrated classifier outputs . A false positive costs , a false negative costs , and correct decisions cost . At what threshold should class be predicted?

53 A fitted logistic model assigns every observation the same probability equal to the sample prevalence. Which property does it have on that same sample?

54 Two predictors use different measurement scales, and ridge regression applies one common penalty parameter without standardizing them. What is the main consequence?

55 When several predictors are strongly correlated and all carry signal, which behavior most clearly distinguishes elastic net from pure lasso?

56 As the ridge penalty parameter increases from while the intercept remains unpenalized, which change is generally expected?

57 Why can lasso produce exactly zero coefficients while ridge generally does not?

58 An ordinary least-squares model with an intercept is evaluated on a separate test set and has . What does this mean?

59 For a nested ordinary least-squares model, adding one predictor increases adjusted precisely when which condition holds?

60 A binary classifier has ROC-AUC on a population with no tied scores. Which interpretation is correct?