Q1. Say, you are interested in predicting body mass index (bmi in kg/m2) with the following outcome model: model <- as.formula(paste('bmi ~ age + sex + race + income + diet + smoking')). The name of the analytic, training, and testing datasets are data.analytic, data.train, and data.test, respectively. Which function can be used to fit 10-fold cross-validation in R?

A. fit <- predict(lm(model, data = data.train), newdata = data.test)
B. fit <- caret::train(model, trControl = trainControl(method = 'cv', number = 10), data = data.train, method = 'lm')
C. fit <- caret::train(model, trControl = trainControl(method = 'cv', number = 10), data = data.analytic, method = 'lm')
D. fit <- caret::train(model, trControl = trainControl(method = 'cv', number = 10), data = data.test, method = 'lm

Q2. Say, 'bmi' is a binary variable in Q1, defined as bmi > 30kg/m2 or not. Which function would you likely use to fit 10-fold cross-validation?

A. fit <- predict(glm(model, data = data.train, family = 'binomial'), newdata = data.test)
B. fit <- caret::train(model, trControl = trainControl(method = 'cv', number = 10), data = data.train, method = 'glm')
C. fit <- caret::train(model, trControl = trainControl(method = 'cv', number = 10, classProbs = TRUE, summaryFunction = twoClassSummary), data = data.analytic, method = 'glm')
D. fit <- caret::train(model, trControl = trainControl(method = 'cv', number = 10), data = data.analytic, method = 'glm)

Q3. Which function would you likely use in Q2 if you are interested in fitting LASSO with 10-fold cross-validation?

A. fit.lasso <- caret::train(model, trControl = trainControl(method = 'cv', number = 10, classProbs = TRUE, summaryFunction = twoClassSummary), data = data.analytic, method = 'glmnet', lambda = 0, tuneGrid = expand.grid(alpha = 0, lambda = 1), verbose = FALSE, metric = 'ROC')
B. fit.lasso <- caret::train(model, trControl = trainControl(method = 'cv', number = 10, classProbs = TRUE, summaryFunction = twoClassSummary), data = data.analytic, method = 'glmnet', lambda = 0, tuneGrid = expand.grid(alpha = seq(0.1, 0.5,by = 0.05), lambda = seq(0.05, 0.9, by = 0.05)), verbose = FALSE, metric = 'ROC')
C. fit.lasso <- caret::train(model, trControl = trainControl(method = 'cv', number = 10, classProbs = TRUE, summaryFunction = twoClassSummary), data = data.analytic, method = 'glmnet', lambda = 0, tuneGrid = expand.grid(alpha = 1, lambda = 0), verbose = FALSE, metric = 'ROC')
D. fit.lasso <- caret::train(model, trControl = trainControl(method = 'cv', number = 10, classProbs = TRUE, summaryFunction = twoClassSummary), data = data.analytic, method = 'glmnet', lambda = 0, tuneGrid = expand.grid(alpha = 0, lambda = 0), verbose = FALSE, metric = 'ROC')

Q4. Which of the following is an ensemble method?

A. fit.ensemble <- caret::train(model, trControl = trainControl(method = 'cv', number = 10, classProbs = TRUE, summaryFunction = twoClassSummary), data = data.analytic, method = 'glm', verbose = FALSE, metric = 'ROC')
B. fit.ensemble <- caret::train(model, trControl = trainControl(method = 'cv', number = 10, classProbs = TRUE, summaryFunction = twoClassSummary), data = data.analytic, method = 'rpart', verbose = FALSE, metric = 'ROC')
C. fit.ensemble <- caret::train(model, trControl = trainControl(method = 'cv', number = 10, classProbs = TRUE, summaryFunction = twoClassSummary), data = data.analytic, method = 'bag', verbose = FALSE, bagControl = bagControl(fit = ldaBag$fit, predict = ldaBag$pred, aggregate = ldaBag$aggregate), metric = 'ROC')
D. fit.lasso <- caret::train(model, trControl = trainControl(method = 'cv', number = 10, classProbs = TRUE, summaryFunction = twoClassSummary), data = data.analytic, method = 'glmnet', lambda = 0, tuneGrid = expand.grid(alpha = 1, lambda = 0), verbose = FALSE, metric = 'ROC')

Q5. Which is an example of unsupervised learning? (Select All that apply)

A. fit.unsupervised <- caret::train(model, trControl = trainControl(method = 'cv', number = 10, classProbs = TRUE, summaryFunction = twoClassSummary), data = data.analytic, method = 'rpart', verbose = FALSE, metric = 'ROC')
B. fit.unsupervised <- varclus(model, data = data.analytic)
C. fit.unsupervised <- kmeans(data.analytic, centers = 2)
D. fit.unsupervised <- glm(model, data = data.analytic, family = binomial)