Goodnessoffit tests are used to compare proportions of levels of a nominal variable to theoretical proportions. Common goodnessoffit tests are Gtest, chisquare, and binomial or multinomial exact tests.
In general, there are no assumptions about the distribution of data for these tests. However, the results of chisquare tests and Gtests can be inaccurate if statistically expected cell counts are low. A rule of thumb is that all statistically expected cell counts should be 5 or greater for chisquare and Gtests. For a more complete discussion, see McDonald in the “Optional Readings” section for details on what constitutes low cell counts.
One approach is to use exact tests, which are not bothered by low cell counts. However, if there are not low cell counts, using Gtest or chisquare test is fine. The advantage of chisquare tests is that your audience may be more familiar with them.
Gtests are also called likelihood ratio tests, and “Likelihood Ratio ChiSquare” by SAS.
Appropriate data
• A nominal variable with two or more levels
• Theoretical, typical, or neutral values for the proportions for this variable are needed for comparison
• Gtest and chisquare tests may not be appropriate if there are cells with low expected counts
Hypotheses
• Null hypothesis: The proportions for the levels for the nominal variable are not different from the theoretical proportions.
• Alternative hypothesis (twosided): The proportions for the levels for the nominal variable are different from the theoretical proportions.
Interpretation
Significant results can be reported as “The proportions for the levels for the nominal variable were statistically different from the theoretical proportions.”
Packages used in this chapter
The packages used in this chapter include:
• EMT
• DescTools
• ggplot2
• rcompanion
The following commands will install these packages if they are not already installed:
if(!require(EMT)){install.packages("EMT")}
if(!require(DescTools)){install.packages("DescTools")}
if(!require(ggplot2)){install.packages("ggplot2")}
if(!require(rcompanion)){install.packages("rcompanion")}
Goodnessoffit tests for nominal variables example
As part of a demographic survey of students in his environmental issues webinar series, Alucard recorded the race and ethnicity of his students. He wants to compare the data for his class to the demographic data for Cumberland County, New Jersey as a whole
Race Alucard’s_class County_proportion
White 20 0.775
Black 9 0.132
American Indian 9 0.012
Asian 1 0.054
Pacific Islander 1 0.002
Two or more races 1 0.025
  
Total 41 1.000
Ethnicity Alucard’s_class County_proportion
Hispanic 7 0.174
Not Hispanic 34 0.826
  
Total 41 1.000
Exact tests for goodnessoffit
Race data
observed = c(20, 9, 9, 1, 1, 1)
theoretical = c(0.775, 0.132, 0.012, 0.054, 0.002, 0.025)
library(EMT)
multinomial.test(observed, theoretical)
### This can take a long time!
Exact Multinomial Test, distance measure: p
Events pObs p.value
1370754 0 0
### A faster, approximate test by Monte
Carlo simulation
observed = c(20, 9, 9, 1, 1, 1)
theoretical = c(0.775, 0.132, 0.012, 0.054, 0.002, 0.025)
library(EMT)
multinomial.test(observed, theoretical,
MonteCarlo = TRUE)
Exact Multinomial Test, distance measure: p
Events pObs p.value
1370754 0 0
Ethnicity data
x = 7
n = 41
theoretical = 0.174
binom.test(x, n, theoretical)
Exact binomial test
number of successes = 7, number of trials = 41, pvalue = 1
Gtest for goodnessoffit
Race data
observed = c(20, 9, 9, 1, 1, 1)
theoretical = c(0.775, 0.132, 0.012, 0.054, 0.002, 0.025)
library(DescTools)
GTest(x = observed,
p = theoretical,
correct="none")
### Correct: "none"
"williams" "yates"
Log likelihood ratio (Gtest) goodness of fit test
data: observed
G = 46.317, Xsquared df = 5, pvalue = 7.827e09
Check expected counts
Test = GTest(x = observed,
p = theoretical,
correct="none")
Test$expected
[1] 31.775 5.412 0.492 2.214 0.082 1.025
### Note the low expected counts: 0.492, 0.082, and
1.025,
### suggesting the test may not be valid.
Ethnicity data
observed = c(7, 34)
theoretical = c(0.174, 0.826)
library(DescTools)
GTest(x = observed,
p = theoretical,
correct = "none")
### Correct: "none"
"williams" "yates"
Log likelihood ratio (Gtest) goodness of fit test
G = 0.0030624, Xsquared df = 1, pvalue = 0.9559
Check expected counts
Test = GTest(x = observed,
p = theoretical,
correct = "none")
Test$expected
[1] 7.134 33.866
### There are no low expected counts.
Chisquare test for goodnessoffit
Race data
observed = c(20, 9, 9, 1, 1, 1)
theoretical = c(0.775, 0.132, 0.012, 0.054, 0.002, 0.025)
chisq.test(x = observed,
p = theoretical)
Chisquared test for given probabilities
Xsquared = 164.81, df = 5, pvalue < 2.2e16
Check expected counts
Test = chisq.test(x = observed,
p = theoretical)
Test$expected
[1] 31.775 5.412 0.492 2.214 0.082 1.025
### Note the low expected counts: 0.492, 0.082, and
1.025,
### suggesting the test may not be valid.
Ethnicity data
observed = c(7, 34)
theoretical = c(0.174, 0.826)
chisq.test(x = observed,
p = theoretical)
Chisquared test for given probabilities
Xsquared = 0.0030472, df = 1, pvalue = 0.956
Check expected counts
Test = chisq.test(x = observed,
p = theoretical)
Test$expected
[1] 7.134 33.866
### There are no low expected counts.
Effect size for goodnessoffit tests
Cramér's V, a measure of association used for 2dimensional contingency tables, can be modified for use in goodnessoffit tests for nominal variables. In this context, a value of Cramér's V of 0 indicates that observed values match theoretical values perfectly. If theoretical proportions are equally distributed, the maximum for value for Cramér's V is 1. However, if theoretical proportions vary among categories, Cramér’s V can exceed 1.
Interpretation of effect sizes necessarily varies by discipline and the expectations of the experiment, but for behavioral studies, the guidelines proposed by Cohen (1988) are sometimes followed. The following guidelines are based on the suggestions of Cohen (2 ≤ k ≤ 6) and an extension of these values (7 ≤ k ≤ 10). They should not be considered universal.
The following table is appropriate for chisquare goodnessoffit tests, and similar tests, where the theoretical proportions are equally distributed among categories. k is the number of categories.

Small

Medium 
Large 
k =2 
0.100 – < 0.300 
0.300 – < 0.500 
≥ 0.500 
k =3 
0.071 – < 0.212 
0.212 – < 0.354 
≥ 0.354 
k =4 
0.058 – < 0.173 
0.173 – < 0.289 
≥ 0.289 
k =5 
0.050 – < 0.150 
0.150 – < 0.250 
≥ 0.250 
k =6 
0.045 – < 0.134 
0.134 – < 0.224 
≥ 0.224 
k =7 
0.043 – < 0.130 
0.130 – < 0.217 
≥ 0.217 
k =8 
0.042 – < 0.127 
0.127 – < 0.212 
≥ 0.212 
k =9 
0.042 – < 0.125 
0.125 – < 0.209 
≥ 0.209 
k =10 
0.041 – < 0.124 
0.124 – < 0.207 
≥ 0.207 
Examples of effect size for goodnessoffit tests
Since the theoretical proportions in these example aren’t equally distributed among categories, these values of Cramér’s V may be difficult to interpret.
Race data
observed = c(20, 9, 9, 1, 1, 1)
theoretical = c(0.775, 0.132, 0.012, 0.054, 0.002, 0.025)
library(rcompanion)
cramerVFit(x = observed,
p = theoretical)
Cramer V
0.8966
Ethnicity data
observed = c(7, 34)
theoretical = c(0.174, 0.826)
cramerVFit (x = observed,
p = theoretical)
Cramer V
0.008621
Simple bar plots
Race example
In order to facilitate plotting with R base graphics, the vector of counts is converted to proportions, and the theoretical and observed proportions are combined into a table, named XT.
See the Introduction to Tests for Nominal Variables chapter for a few additional options for how bar plots are plotted.
observed = c(20, 9, 9, 1, 1, 1)
theoretical = c(0.775, 0.132, 0.012, 0.054, 0.002, 0.025)
Observed.prop = observed / sum(observed)
Theoretical.prop = theoretical
Observed.prop = round(Observed.prop, 3)
Theoretical.prop = round(Theoretical.prop, 3)
XT = rbind(Theoretical.prop, Observed.prop)
colnames(XT) = c("White", "Black", "AI",
"Asian",
"PI", "Two+")
XT
White Black AI Asian PI Two+
Theoretical.prop 0.775 0.132 0.012 0.054 0.002 0.025
Observed.prop 0.488 0.220 0.220 0.024 0.024 0.024
barplot(XT,
beside = T,
xlab = "Race",
col = c("cornflowerblue","blue"),
legend = rownames(XT))
In order to plot with ggplot2, the table XT will be converted into a data frame in long format.
Race = c(colnames(XT), colnames(XT))
Race = factor(Race,
levels=unique(colnames(XT)))
Proportion = c(rep("Theoretical", 6),
rep("Observed", 6))
Value = c(XT[1,], XT[2,])
DF = data.frame(Race, Proportion, Value)
DF
Race Proportion Value
1 White Theoretical 0.775
2 Black Theoretical 0.132
3 AI Theoretical 0.012
4 Asian Theoretical 0.054
5 PI Theoretical 0.002
6 Two+ Theoretical 0.025
7 White Observed 0.488
8 Black Observed 0.220
9 AI Observed 0.220
10 Asian Observed 0.024
11 PI Observed 0.024
12 Two+ Observed 0.024
library(ggplot2)
ggplot(DF,
aes(x = Race, y = Value, fill = Proportion)) +
geom_bar(stat = 'identity', position = 'dodge') +
scale_fill_manual(values = c("cornflowerblue",
"blue")) +
ylab("Proportion")
Ethnicity example
observed = c(7, 34)
theoretical = c(0.174, 0.826)
Observed.prop = observed / sum(observed)
Theoretical.prop = theoretical
Observed.prop = round(Observed.prop, 3)
Theoretical.prop = round(Theoretical.prop, 3)
XT = rbind(Theoretical.prop, Observed.prop)
colnames(XT) = c("Hispanic", "Not hispanic")
XT
Hispanic Not hispanic
Theoretical.prop 0.174 0.826
Observed.prop 0.171 0.829
barplot(XT,
beside = T,
xlab = "Ethnicity",
col = c("firebrick1","firebrick"),
legend = rownames(XT),
ylim = c(0, 1.4))
Ethnicity = c(colnames(XT), colnames(XT))
Ethnicity = factor(Ethnicity,
levels=unique(colnames(XT)))
Proportion = c(rep("Theoretical", 2), rep("Observed", 2))
Value = c(XT[1,], XT[2,])
DF = data.frame(Ethnicity, Proportion, Value)
DF
Ethnicity Proportion Value
1 Hispanic Theoretical 0.174
2 Not hispanic Theoretical 0.826
3 Hispanic Observed 0.171
4 Not hispanic Observed 0.829
library(ggplot2)
ggplot(DF,
aes(x = Ethnicity, y = Value, fill = Proportion)) +
geom_bar(stat = 'identity', position = 'dodge') +
scale_fill_manual(values =
c("firebrick1","firebrick")) +
ylab("Proportion")
Optional: Multinomial test example with plot and confidence intervals
This is an example of a multinomial test that includes a bar plot showing confidence intervals. The data is a simple vector of counts.
For a similar example using twoway count data which is organized into a data frame, see the "Examples of basic plots for nominal data" section in the Basic Plots chapter.
Walking to the store, Jerry Coyne observed the colors of nail polish on women’s toes (Coyne, 2016). Presumably because that’s the kind of thing retired professors are apt to do. He concluded that red was a more popular color but didn’t do any statistical analysis to support his conclusion.
Color of polish Count
Red 19
None or clear 3
White 1
Green 1
Purple 2
Blue 2
We will use a multinomial goodness of fit test to determine if there is an overall difference in the proportion of colors (multinomial.test function in the EMT package). The confidence intervals for each proportion can be found with the MultinomCI function in the DescTools package. The data then needs to be manipulated some so that we can plot the data as counts and not proportions.
Note here that the theoretical counts are simply 1 divided by the number of treatments. In this case the null hypothesis is that the observed proportions are all the same. In the examples above, the null hypothesis was that the observed proportions were not different than theoretical proportions. It’s two ways to think of the same null hypothesis.
The confidence intervals for each proportion can be used as a posthoc test, to determine which proportions differ from each other, or to determine which proportions differ from 0.
nail.color = c("Red", "None", "White",
"Green", "Purple", "Blue")
observed = c( 19, 3, 1, 1, 2, 2 )
theoretical = c( 1/6, 1/6, 1/6, 1/6, 1/6, 1/6 )
library(EMT)
multinomial.test(observed,
theoretical)
### This may take a while. Use Monte Carlo for
large numbers.
Exact Multinomial Test, distance measure: p
Events pObs p.value
237336 0 0
library(rcompanion)
cramerVFit(observed)
### Assumes equal proportions for theoretical
values.
Cramer V
0.6178
library(DescTools)
MCI = MultinomCI(observed,
conf.level=0.95,
method="sisonglaz")
MCI
est lwr.ci upr.ci
[1,] 0.67857143 0.5357143 0.8423162
[2,] 0.10714286 0.0000000 0.2708876
[3,] 0.03571429 0.0000000 0.1994590
[4,] 0.03571429 0.0000000 0.1994590
[5,] 0.07142857 0.0000000 0.2351733
[6,] 0.07142857 0.0000000 0.2351733
### Order the levels, otherwise R will
alphabetize them
Nail.color = factor(nail.color,
levels=unique(nail.color))
### For plot, Create variables of counts, and then
wrap them into a data frame
Total = sum(observed)
Count = observed
Lower = MCI[,'lwr.ci'] * Total
Upper = MCI[,'upr.ci'] * Total
Data = data.frame(Count, Lower, Upper)
Data
Count Lower Upper
1 19 15 23.584853
2 3 0 7.584853
3 1 0 5.584853
4 1 0 5.584853
5 2 0 6.584853
6 2 0 6.584853
library(ggplot2)
ggplot(Data, ### The data frame to
use.
aes(x = Nail.color,
y = Count)) +
geom_bar(stat = "identity",
color = "black",
fill = "gray50",
width = 0.7) +
geom_errorbar(aes(ymin = Lower,
ymax = Upper),
width = 0.2,
size = 0.7,
position = pd,
color = "black"
) +
theme_bw() +
theme(axis.title = element_text(face = "bold")) +
ylab("Count of observations") +
xlab("Nail color")
Bar plot of the count of the color of women’s toenail
polish observed by Jerry Coyne while walking to the store. Error bars indicate
95% confidence intervals (Sison and Glaz method).
Optional readings
“Small numbers in chisquare and G–tests” in McDonald, J.H. 2014. Handbook of Biological Statistics. www.biostathandbook.com/small.html.
References
Cohen, J. 1988. Statistical Power Analysis for the Behavioral Sciences, 2nd Edition. Routledge.
Coyne, J. A. 2016. Why is red nail polish so popular? Why Evolution is True. whyevolutionistrue.wordpress.com/2016/07/02/whyisrednailpolishsopopular/.
“Chisquare Test of GoodnessofFit” in Mangiafico, S.S. 2015a. An R Companion for the Handbook of Biological Statistics, version 1.09. rcompanion.org/rcompanion/b_03.html.
“Exact Test of GoodnessofFit” in Mangiafico, S.S. 2015a. An R Companion for the Handbook of Biological Statistics, version 1.09. rcompanion.org/rcompanion/b_01.html.
“G–test of GoodnessofFit” in Mangiafico, S.S. 2015a. An R Companion for the Handbook of Biological Statistics, version 1.09. rcompanion.org/rcompanion/b_04.html.
“Repeated G–tests of GoodnessofFit” in Mangiafico, S.S. 2015a. An R Companion for the Handbook of Biological Statistics, version 1.09. rcompanion.org/rcompanion/b_09.html.
Exercises Beta
1. What do you conclude about whether the race of Alucard’s
students match the proportions in the county? Include, as appropriate:
• The statistical test you are using, pvalue from that test, conclusion of the test, and any other comments on the test.
• The effect size and interpretation, or perhaps why you don’t want to rely on that information.
• Description of the observed counts or proportions relative to the theoretical proportions. Is there anything notable?
• Any other practical conclusions you wish to note.
2. What do you conclude about whether the ethnicity of
Alucard’s students match the proportions in the county? Include, as
appropriate:
• The statistical test you are using, pvalue from that test, conclusion of the test, and any other comments on the test.
• The effect size and interpretation, or perhaps why you don’t want to rely on that information.
• Description of the observed counts or proportions relative to the theoretical proportions. Is there anything notable?
• Any other practical conclusions you wish to note.