C:/R/functions txt



Download 31.52 Kb.
Date06.08.2017
Size31.52 Kb.
#27994
R Scripts for Chapter 7
# Script 7.1 - First example

source("C:/R/functions.txt")

library(car)
# Reading the data

chap7.ex1 <- read.table("c:/rbook/chap6.ex1.txt", header = TRUE)

attach(chap7.ex1)
# Drawing Figure 7.5

plot(jitter(wkdaded, factor = 1), jitter(wkmomed, factor = 1),

pch = 20, xlim = c(0, 10), ylim = c(0, 10))
# Drawing Figure 7.6

plot(jitter(wkdaded,factor=1),jitter(wkmomed,factor=1),pch=20,xlim=c(0,10),ylim=c(0,10))

abline(3.0,.5)

abline(0,1)

text(2,7, "Line 1")

text(8,4, "Line 2")

arrows(1.9, 6.2, 1.9, 4, code = 2, length = .15, angle = 20)

arrows(8, 4.5, 8, 6.7, code = 2, length = .15, angle = 20)


# Finding the regression equation

reg <- lm(wkmomed~wkdaded)

reg
# Adding the predicted values and residuals to the data frame

chap7.ex1$pred <- fitted(reg)

chap7.ex1$res <- resid(reg)

names(chap7.ex1)

attach(chap7.ex1)

pred


res
# Looking at the intercorrelations

cor.mat <- cov2cor(cov(chap7.ex1))

print(cor.mat, digits = 3)
# Reversing the IV and DV

reg.rev <- lm(wkdaded~wkmomed)

reg.rev
# Plotting the confidence and prediction intervals: Figure 7.10

interval.frame <- data.frame(wkdaded = 0:10)

pp <- predict(reg, int = "p", newdata = interval.frame)

pc <- predict(reg, int = "c", newdata = interval.frame)

plot(jitter(wkdaded), jitter(wkmomed), ylim = c(0, 10),

xlim = c(0, 10), pch = 16)

pred.wkmomed <- interval.frame$wkdaded

matlines(pred.wkmomed, pc, lty = c(1, 2, 2), col = "black")

matlines(pred.wkmomed, pp, lty = c(1, 3, 3), col = "black")
chap7.ex1$z.wkmomed <- scale(wkmomed)

chap7.ex1$z.wkdaded <- scale(wkdaded)

attach(chap7.ex1)
z.reg <- lm(z.wkmomed~z.wkdaded)

z.reg
# Script 7.2 - Plotting Figure 7.7, the SSE as a function of intercept and slope

# Clean out workspace

rm(list = ls())

library(MASS)

library(lattice)

errors <- read.table("c:/rbook/sse.txt", header = TRUE)

names(errors)

attach(errors)

wireframe(sse ~ b0*b1, scales = list(arrows=FALSE,

cex = .45,

font = 3, tck = 1),

screen = list(z = -245, x = -50, y = 0),shade = FALSE, aspect = c(87/87, 1.0))

# Script 7.3 - Complete example with Males and Females

# Clear out workspace

rm(list = ls())

library(car)

source("C:/R/functions.txt")


# Read in the data

ecls200 <- read.table("c:/rbook/ecls200.txt", header = TRUE)

attach(ecls200)

names(ecls200)


# Create factor variable with labels

ecls200$f.gender <- factor(gender, levels = 1:2, labels =

c("Male", "Female"), ordered = TRUE)
# Create separate data frames by gender

ecls200.female <- subset(ecls200, f.gender == "Female")

ecls200.male <- subset(ecls200, f.gender == "Male")

attach(ecls200.male)

attach(ecls200.female)

attach(ecls200)


# Looking at General Knowledge measured Fall-K

tapply(c1rgscal, f.gender, length)

tapply(c1rgscal, f.gender, table)

tapply(c1rgscal, f.gender, summary)

tapply(c1rgscal, f.gender, sd)

tapply(c1rgscal, f.gender, skewness)

tapply(c1rgscal, f.gender, SEsk)

tapply(c1rgscal, f.gender, kurtosis)

tapply(c1rgscal, f.gender, SEku)

boxplot(c1rgscal ~ f.gender, ylab = "Fall-K General Knowledge Scores")


# Looking at Reading Scores measured Spring-1

tapply(c4rrscal, f.gender, length)

tapply(c4rrscal, f.gender, table)

tapply(c4rrscal, f.gender, summary)

tapply(c4rrscal, f.gender, sd)

tapply(c4rrscal, f.gender, skewness)

tapply(c4rrscal, f.gender, SEsk)

tapply(c4rrscal, f.gender, kurtosis)

tapply(c4rrscal, f.gender, SEku)

boxplot(c4rrscal ~ f.gender, ylab = "Spring-1 Reading Scores")

# Produce a scatterplot with subgroups

# Set plotting character to 1 for all cases

pchval <- rep(1, length = length(f.gender))

# Set plotting character to 18 for Males

pchval[f.gender == "Male"] <- 18
plot(jitter(c1rgscal, factor = 1), jitter(c4rrscal, factor = 1),

pch = pchval)

legend(32, 28, c("Male", "Female"), pch = c(18, 1))

legend(32, 40, c("Male", "Female"), lty = c(1, 2))

with(ecls200.male,abline(lm(c4rrscal ~ c1rgscal), lty = 1))

with(ecls200.female,abline(lm(c4rrscal~c1rgscal), lty = 2))


attach(ecls200.female)

mod.female <- lm (c4rrscal ~ c1rgscal)

mod.female

ecls200.female$res <- resid(mod.female)

attach(ecls200.female)

ecls200.female$z.res <- scale(res)

attach(ecls200.female)

range(z.res)


# Look at residuals)

plot(z.res, ylim = c(-5, 5))

abline(h=2.5, lty = 2)

abline(h= - 2.5, lty = 2)

which(z.res < -2)

which(z.res > 2)


# Look at Cook's D

# cook.d() function updated to cooks.distance()

denom <- length(res) - 2

plot(cooks.distance(mod.female))

abline(h = 4/denom, lty = 2)

which(cooks.distance(mod.female) > (4/denom))

detach(ecls200.female)
attach(ecls200.male)

mod.male <- lm (c4rrscal ~ c1rgscal)

mod.male

ecls200.male$res <- resid(mod.male)

attach(ecls200.male)

ecls200.male$z.res <- scale(res)

attach(ecls200.male)

range(z.res)


# Look at residuals)

plot(z.res, ylim = c(-5, 5))

abline(h = 2.5, lty = 2)

abline(h = - 2.5, lty = 2)

which(z.res < -2)

which(z.res > 2)


# Look at Cook's D

denom <- length(res) - 2

plot(cooks.distance(mod.male))

abline(h=4/denom, lty = 2)

which(cooks.distance(mod.male) > (4/denom))

detach(ecls200.male)


# Looking at the two groups as one group

attach(ecls200)

mod.ecls200 <- lm (c4rrscal ~ c1rgscal)

mod.ecls200

ecls200$res <- resid(mod.ecls200)

attach(ecls200)

ecls200$z.res <- scale(res)

attach(ecls200)

range(z.res)
# Look at residuals

plot(z.res, ylim = c(-5,5))

abline(h = 2.5, lty = 2)

abline(h= - 2.5, lty = 2)

which(z.res < -2)

which(z.res > 2)


# Look at Cook's D

denom <- length(res) - 2

plot(cooks.distance(mod.ecls200))

abline(h = 4/denom, lty = 2)

which(cooks.distance(mod.ecls200) > (4/denom))
# Script 7.4 - Solution to Exercises

rm(list = ls())

library(car)

source("C:/R/functions.txt")


# Read the data from Exercise 6.3

ex7.3 <- read.table("c:/rbook/chap6.ex3.txt", header = TRUE)

attach(ex7.3)
# Calculate predicted values

ex7.3$yhat <- -5.69 + .0677*iq

attach(ex7.3)
# Estimate errors

ex7.3$error <- gpa - yhat

attach(ex7.3)
# Find squared errors

ex7.3$error.sq <- error^2

attach(ex7.3)
# Examining the results

ex7.3
# Finding the sums

sum(gpa)

sum(yhat)

sum(error)

sum(error.sq)


# Using lm function in R to do analysis

model <- lm(gpa ~ iq)

summary(model)

anova(model)




Download 31.52 Kb.

Share with your friends:




The database is protected by copyright ©ininet.org 2024
send message

    Main page