R Parameterized Modeling

Question 1

Omit Missing Values From Data Frame

Answer 1

na.omit(DataFrame)

Question 2

how to count missing values in a column

Answer 2

sum(is.na(DataFrame$column))

Question 3

How to perform Best Subset Selection

Answer 3

"library(leaps)
regfit.full = regsubsets(Salary~., Hitters)
#by default, this only tests up to the best eight variable model, 
#you can override this by using the nvmax option:
regfit.full = regsubsets(Salary~., data = Hitters, nvmax = 19)"

Question 4

How to interpret output of best subset model

Answer 4

It will return the best model for models with n variables. For example, for two variable model it will mark the appropriate two columns with an asterik

Question 5

how to plot best subset using best subset selection

Answer 5

“plot(regfit.full, scale = ““adjr2””)

see ?plot.regsubsets for help”

Question 6

Do forward/backward stepwise selection

Answer 6

“use method parameter in regsubsets function

regfit. fwd = regsubsets(Salary~., data= Hitters, method=”“forward””)
regfit. bwd = regsubests(Salary~., data = Hitters, method = ‘backward’)”

Question 7

Create a training/test set

Answer 7

“set.seed(1)
train = sample(c(TRUE, FALSE), nrow(Hitters), rep = TRUE)
test = (!train)
regfit.best = regsubsets(Salary~., data = Hitters[train, ])”

Question 8

Create a matrix of X’s, which is a data structure used by many regression packages

Answer 8

“model.matrix(Formula, Dataframe)
this function creates a matrix with an intercept of 1, and dummy codes factor vars for you. Used by many regression functions behind the scenes to train and predict values. You must use this function incase some function does not have a predict function (like regsubsets)
remember - Formula = Y~., “

Question 9

What should you type in the R console to install the “car” package?

Answer 9

install.packages(‘psych’)

Question 10

What should you type to create a matrix “a” comprising all natural numbers from 1 to 10, with 2 rows and 5 columns.

Answer 10

a = matrix(1:10, 2, 5)

Question 11

PCR in R

Answer 11

library(pls)

set. seed(2)
pcr. fit = pcr(Salary~., data = Hitters, scale = TRUE, validation = ‘CV’)

Question 12

PLS in R

Answer 12

library(pls)
set.seed(1)
pls.fit=plsr(Salary~., data=Hitters, subset = train, scale = TRUE, validation = ‘CV’)
summary(pls.fit)

Question 13

PLS vs. PCR

Answer 13

PCR tries to maximize variance explained by the predictors (the principal components), whereas PLS searches for predictors that explain variance in BOTH the predictors and the response - usually PLS explains more variance with fewer predictors.

Question 14

polynomial regression

Answer 14

fit = lm(wage~poly(age, 4), data=Wage)