# R code to analyze an r by c contingency table
# and do chi-squared test of independence

# We look at the snoring / heart disease data set:

# The observed counts may be input as a matrix as follows:

snore.heart.data <- matrix(c(24,35,51,1355,603,416), nrow=2, ncol=3, byrow=TRUE, 
dimnames = list(c("Yes", "No"), c("Never", "Occasionally", "~ Every Night")))

# Printing the 2 by 3 table:

as.table(snore.heart.data)

# We test for independence of the two classifications
# using the chisq.test() function in R:

chisq.test(snore.heart.data, correct=FALSE)

# The P-value is near zero.

# A quick way to get expected cell counts to verify large-sample assumption is met:

expected.counts <- (apply(snore.heart.data,1,sum) %o% apply(snore.heart.data,2,sum))/sum(snore.heart.data)
print(expected.counts)


####################################################################


# R code to analyze an r by c contingency table
# and do chi-squared test of homogeneity

# We look at the voter survey data set:

# The observed counts may be input as a matrix as follows:

voter.data <- matrix(c(61,50,39,56,40,54,47,37,66), nrow=3, ncol=3, byrow=TRUE, 
dimnames = list(c("Upstate", "Midlands", "Coastal"), c("Approve", "Neutral", "Disapprove")))

# Printing the 3 by 3 table:

as.table(voter.data)

# We test for homogeneity
# using the chisq.test() function in R:

chisq.test(voter.data, correct=FALSE)

# The P-value is 0.027.

# A quick way to get expected cell counts to verify large-sample assumption is met:

expected.counts <- (apply(voter.data,1,sum) %o% apply(voter.data,2,sum))/sum(voter.data)
print(expected.counts)

####################################################################


# R code to do Fisher's exact test
# on a made-up small data set 
# (3 by 3 table)

small.data <- matrix(c(6,5,3,5,4,5,4,3,6), nrow=3, ncol=3, byrow=TRUE, 
dimnames = list(c("Upstate", "Midlands", "Coastal"), c("Approve", "Neutral", "Disapprove")))

# Printing the 3 by 3 table:

as.table(small.data)

fisher.test(small.data)