Scribd
Upload
47 views2 pages

R Code For Testing Goodness of Fit, Independence and Homogeneity

This document provides R code examples for three types of categorical data analysis tests: goodness of fit, independence, and homogeneity. For goodness of fit, a chi-squared test is used to determine if the observed frequencies of rose colors match the expected frequencies. For independence, a chi-squared test on a contingency table tests if hair and eye color are associated. For homogeneity, a chi-squared test (with a permutation approach) determines if egg preference proportions are the same between gender groups.

Uploaded by

Pedro Humberto Almeida Gonzaga
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
47 views2 pages

R Code For Testing Goodness of Fit, Independence and Homogeneity

This document provides R code examples for three types of categorical data analysis tests: goodness of fit, independence, and homogeneity. For goodness of fit, a chi-squared test is used to determine if the observed frequencies of rose colors match the expected frequencies. For independence, a chi-squared test on a contingency table tests if hair and eye color are associated. For homogeneity, a chi-squared test (with a permutation approach) determines if egg preference proportions are the same between gender groups.

Uploaded by

Pedro Humberto Almeida Gonzaga
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Categorical Data Analysis Prof.

Klingenberg

R code for testing Goodness of Fit, Independence and Homogeneity


Goodness of Fit:

Example: (Roses) When crossing certain types of red and white roses, one obtains red, white and pink
roses. Theory predicts that the proportion of red to white to pink roses is like 3:2:2. Test the plausibility
of this theory when out of a sample of 80 crosses, 35 are red, 31 are white and 14 are pink. (Note:
Sampling design is multinomial sampling of one variable and we test to see if the multinomial
probabilities are equal to some specified values)

> chisq.test(c(35,31,14), p=c(3,2,2)/7)

Chi-squared test for given probabilities

data: c(35, 31, 14)


X-squared = 6.3479, df = 2, p-value = 0.04184

Conclusion: At a 5% significance level, the data provide sufficient evidence (P-value = 0.0418) that the
proportion of red to white to pink roses is different from 3:2:2.

Independence:

Example: (Hair and Eye Color) In a sample of 65 students, we recorded the hair color (categories blond,
brown, dark) and eye color (categories bright, dark). The table below summarizes the counts. Null
hypothesis: Hair and Eye color are independent. Alternative Hypothesis: Hair and eye color are
associated. (Note: Sampling design is multinomial, where two categorical responses are recorded.)

> table1=matrix(c(12,2,8,25,6,12),ncol=3)
> colnames(table1)=c("blond","brown","dark")
> rownames(table1)=c("bright","dark")
> table1
blond brown dark
bright 12 8 6
dark 2 25 12

> chisq.test(table1)

Pearson's Chi-squared test

data: table1
X-squared = 15.938, df = 2, p-value = 0.000346

Conclusion: There is sufficient evidence (P-value=0.0003) that hair and eye color of students are
associated.
Homogeneity (Comparing proportions across two groups):

Example : (Egg) We asked n1=25 females (group 1) and n2=17 males (group 2) how they preferred their
Sunday morning breakfast egg (Sunny Side Up, Over Easy or Scrambled). The data are summarized in the
table below. Is the distribution of egg preference the same for males and females? (Are the proportions
homogeneous across the two groups?) Null hypothesis: distributions are the same for females and
males; Alternative Hypothesis: Distributions are not the same (Note: Sampling design is product
multi(=bi) nomial.)

> table2=matrix(c(5,9,12,3,7,5),ncol=3)
> colnames(table2)=c("Sunny","Over Easy","Scrambled")
> rownames(table2)=c("Females","Males")
> table2
Sunny Over Easy Scrambled
Females 5 12 7
Males 9 3 5
> chisq.test(table2)

Pearson's Chi-squared test

data: table2
X-squared = 5.8516, df = 2, p-value = 0.05362

Warning message:
In chisq.test(table2) : Chi-squared approximation may be incorrect

When we don’t trust the validity of the asymptotic approximation (see Warning), the permutation
approach we discussed in class is actually implemented in the chisq.test() function and safer to use here
(you can also use it for the previous example about hair and eye color):

> chisq.test(table2,simulate.p.value = TRUE, B = 10000)

Pearson's Chi-squared test with simulated p-value (based on


10000
replicates)

data: table2
X-squared = 5.8516, df = NA, p-value = 0.0611

Conclusion: We have insufficient evidence (P-value = 0.0611) to conclude that the distribution of egg
preference is different for females and males. (Note: The exact test here is based on the proportion of
sampled tables (with the same margin) with X2 statistics at least as large or larger than the observed
test statistic. Another option is to use fisher.test() which gets the exact P-value using the exact null table
probabilities.)

You might also like