DSE R CODE

Question 1

reading data

Answer 1

Advertising = read.csv(“Data/Advertising.csv”, head = TRUE)
head(Advertising)

Question 2

Generate lm model

Answer 2

lm1 = lm(sales ~ TV, data = Advertising)

Question 3

Generate r output table

Answer 3

summary(lm1)

Question 4

Gnerate confidence interval for coeff of variable and constnat (95% or 90%)

Answer 4

confint(lm1)
confint(lm1,level=0.9)

Question 5

How to plot scatter plot? Sales against tv

label axis
colour red
dots
with regression line line width 3 and colour boue

Answer 5

plot(x = Advertising$TV, y = Advertising$sales,
xlab = “TV”, ylab = “Sales”, col = “red”, pch = 19)
abline(lm1, col = “blue”, lwd = 3)

Question 6

How to get coefficient from R output table or summary table?>

Answer 6

summary(lm4)$coefficients

Question 7

How to determine whether there is relationship between variables? 4dp

Answer 7

round(cor(Advertising), digits = 4)

Question 8

how to get adjusted r squared?

Answer 8

summary(lm_mpg1)$adj.r.squared

Question 9

How to load data from ISLR package?
Get auto fata

Answer 9

library(ISLR)
data(Auto)

Question 10

how to generate residuals plot( such as q-q plot) ?

Answer 10

par(mfrow = c(2, 2))
plot(lm_mpg1)

Question 11

How to generate scale-location plot?

Answer 11

plot(lm_mpg2, which = 3)

Question 12

xclude name variable from linear model (mpg= everything except name)

Answer 12

lm_mpg4 = lm(mpg ~ . - name, data = Auto)

Question 13

how to add abels to nominal categorical data?

Answer 13

Auto$origin = factor(Auto$origin, labels = c(“American”, “European”, “Japanese”))

Question 14

how to generate logistic model ?

Answer 14

glm_fit = glm(default ~ balance, data = Default, family = binomial)

need to put familiy=binmomial

BINOMIAL IS NOT A STRING!!!!!

Question 15

How to have predicted probabilities from your own data

Answer 15

df_new = data.frame(student = c(“Yes”, “No”),
balance = c(1500, 1500), income = c(40000, 40000))

predict(glm_fit, newdata = df_new, type = “response”)

Question 16

How to generate confusion matrix from iyself? START FROM PREDICTION STEP

Answer 16

STEP1: PREDICT FIRST
glm_prob = predict(glm_fit, type = “response”)

STEP 2: TABLE GENERATION
table(glm_prob >= 0.5, Default$default)

Question 17

How to plot ROC curve and AUC? (Start from prediction)

CURVE: LINEWIDTH 3
COLOUR BLACK

PASTE TEXT
Add line at 0 slope 1 ,dotted , line width 1

Answer 17

STEP 1: PREPARE DATA
pred = prediction(glm_prob, Default$default) #performed on training data

perf = performance(pred, measure = “tpr”, x.measure = “fpr”)

auc_perf = performance(pred, measure = “auc”)@y.values[[1]]

STEP 2: Plot ROC curve
plot(perf, lwd = 3, col = “black”)

abline(0, 1, lwd = 1, lty = 2) # add dashed diagonal line

text(0.4, 0.8, paste(“AUC =”, round(auc_perf, 2))) # add text

Question 18

How to generate validation set?

Answer 18

set.seed(1101)

train = sample(10000, 5000, replace = FALSE)

Default_train = Default[train, ]

Default_test = Default[-train, ]

Question 19

How to generate confusion matrix from training set?

Answer 19

glm_prob_train = predict(glm_fit2, type = “response”)

table(glm_prob_train > 0.5, Default_train$default)

Question 20

How to get trainign error after generating train results?

Answer 20

glm_pred = ifelse(glm_prob_train > 0.5, “Yes”, “No”)

mean(glm_pred != Default_train$default)

Question 21

How to plot residual vs leverage plot?

Answer 21

plot(lm1, which =5 )

which =5!!!! not which =4 (cook distance)

Question 22

What is the syntax for knn?

Answer 22

library(kknn) #not knn

fit=kknn(model, train, test, k, kernel=“rectangular”)

Question 23

How to choose not to standardize variable for knn

Answer 23

fit=kknn(model, train, test, k, kernel=“rectangular”, scale =FALSE)

Question 24

How to generate naive bayes model

Answer 24

library(es1071)
fit = naiveBayes(model, data)

Question 25

vWhat assumption does the naive bayes function in r rely on ? (slide 19)

Answer 25

assumption of Normal distribution for quantitative predictors.

Question 26

Do you tune parameters to select for naive bayes in R? (slide 19)

Answer 26

no

Question 27

What is the tree basic syntax?

Answer 27

tree.fit=tree(y~x, data, tree.control)

Question 28

What is the rpart basic syntax?

Answer 28

tree.fit=rpart(y~x, data, rpart.control)

Question 29

What is the syntax for kmenas? Explain the parameters

Answer 29

kmeans(x, centers=k, nstart=n) “centers” specifies K, “nstart” tells R how many random initializations we want to perform.

Question 30

What is the function for pca?

Answer 30

prcomp(x, scale=TRUE) scale means standardization

Question 31

What is the syntax of pcr? What to do if we want to use LOOCV?

Answer 31

fit = pcr(y~ ., data = x, scale = TRUE, validation = “CV”) validation=CV” selects M by 10-fold CV. Use “LOO” for LOOCV.

Question 32

How to plot CV MSE for pcr?

Answer 32

validationplot(fit, val.type=“MSEP”)

Question 33

How to predict on valuews on pcr?

Answer 33

pred = predict(fit, x[test,], ncomp = M)