ASSOCIATION COEFFICIENT" " of Yule

CONTINGENCY TABLE
The Yule 'Q' Coefficient is used to determine if two variables VARIABLE Y
measures on nominal or ordinal scales and ordered tables of
2x2 independent sons. The coefficient was developed and
published in 1912 by the British statistician George Undy Yule
(1871–1951), who named her Q in honor of the Belgian statistician.
Quetelet (1796-1874)
The symmetric association coefficient Q of Yule is defined as:

George Udny Yule (1871 in Morham,

Scotland - 1951 in Cambridge,
England.
PROPERTIES
George was born in Beech Hill, a house in
It is limited to:-1 ≤ ≤
If the variables X and Y are independent, Morham, near Haddington in Scotland.
When she turned sixteen, she entered the
then =
Yes = , it is verified that 12. 21= 0y University College London to study
there is a complete association between the una ingeniería.
variables (attributes) X and Y. Later, he studied physics and subsequently Karl Pearson offered him
Yes = − it is verified that 11. 22= 0y a researcher position at University College London.
there is a complete dissociation between the His first article on statistics appeared in 1895 "On the
variables (attributes) X and Y. correlation of total pauperism with proportion of out-relief
If > , the association is positive. where I introduced the application of correlation coefficients
the study of double entry tables, and it is in 1912 that he publishes
Yes < , the association is negative.
On the Methods of Measuring Association Between Two
"Attributes" in the Journal of the Royal Statistical Society.
Interpretation

The value of Q is between -1 and 1. Its interpretation is as follows:

IfQ = 0There is no association between the modalities or levels of the variables, that is, they are independent.
YesQ = +1, there is a perfect positive association.
YesQ = -1, there is perfect negative association.
The interpretation of the results can be difficult and arbitrary with measures of association. A general rule (Knoke and
Bohrnstedt, 1991) is as follows:

DISADVANTAGES

One problem with Q is that the presence of a

zero in any cell or box causes the
the final quotient has a value of 1 or -1. For
example:

The calculation of Q results in:

− ( 25 )( 45-) (0)(25) 1125

= = = =1
+ ( 25 )( 45+) (0)(35) 1125

Despite Q resulting in 1, the relationship is far from perfect. There were 35 high-income respondents with low participation.
a contradiction with the hypothesis. Another measure, the phi coefficient , does not share that particular characteristic with Q,
For this reason, sometimes it is preferred to use coefficient despite a slightly more complicated calculation.
EXAMPLE

We can ask the question: Is there a relationship between sex and the habit of smoking cigarettes? The data from 100 men.
and 100 women, are the following:

Men Women
If he/she smokes 60 25
Does not smoke 40 75

First, it is studied whether X and Y are statistically independent, if

thus all the coefficients that are requested in the problem would be
equal to zero. For them, the necessary and sufficient condition is tested.
statistical independence

. .
= × = 1,2… ℎ ; = 1,2… ,

Let's look at the first case:

11 .
×
.
Men Total Women
If he/she smokes 60 11 25 12 85 1.
60 85 100
200 200
× 200
Does not smoke 40 21 75 22 115 2.
0.3 0.425 × 0.5 Total 100 .1 100 .1 200
0.3 0.2125

As observed, it does not hold for the first case analyzed.(0.3 ≠ 0.2125)therefore independence is not fulfilled
statistics, and the requested coefficients will have non-null values.
Applying Yule's Q formula, we obtain:

60×7-20×40 3500 3500

= 60×75+25×40
= 4500+1000
= 5500
= 0.63

INTERPRETATION

The Q coefficient of Yule, = 0.63indicates a moderate dependency relationship between sex and smoking habit
cigarettes. Furthermore, since this association is positive = 0.63 is greater than 0,that is to say, men are associated with the habit of
smoking (60) and women's habits of not smoking (75).
 NATIONAL UNIVERSITY OF SAN AGUSTIN OF AREQUIPA
PROFESSIONAL SCHOOL OF PSYCHOLOGY
CURSO: PSICOESTADISTICA PROF.: LICENSED LUIS GUERRA JORDAN
FOURTH WORKSHOP PART II ASSOCIATION COEFFICIENTS × Q OF YULE

In a study to find out if tall men tend to marry tall partners, the following information was published
about the wives of 134 tall husbands and 116 short husbands. Find the coefficient of association between the
height of wives and husbands.

Tall spouses Low partners

High cuffs 112 26
Low handcuffs 22 90

a. Complete the contingency table by number of people (simple percentage frequency) considering the variable
independent X (Level of education) and the dependent variable Y (Level of attention they pay to news, issues
or events that happen in other countries).

Height of Height of the Spouses Height of Height of the Spouses

the wives Highs lows Total the handcuffs highs lows Total
Altas High
Drops Drops
Total Total (%) 100% 100% 100%

b. Create grouped bar charts by absolute frequencies (number of spouses) and by percentages (percentage)
of wives) by percentages considering the height of the parents on the horizontal axis.

c. Carry out the necessary calculations to prove that the necessary and sufficient condition of independence is not met.
statistics
. .
== × = 1,2… ℎ ; = 1,2… ,

d. Calculate and interpret the Q coefficient of Yule.

In a company with 500 workers, with higher education or not, and who work in offices or production lines.
As shown in the following contingency table. Analyze whether there is an association between the two variables using a
respective measure of association.

Chain of
Offices
production
Studies
270 30
superiors
No studies
80 120
superiors

a. Complete the contingency table by number of people (simple percentage frequency) considering the variable
independent X (Level of education) and the dependent variable Y (Level of attention they pay to the news,
issues or events that occur in other countries.
Type of work Type of work
Studies Chain offices Total offices Chains Total
Studies
of of
production production
Superiors Superiors 100%
No No 100%
superiors superiors
Total

b. Create grouped bar charts by absolute frequencies and by percentages (type of work)
considering the type of studies on the horizontal axis.

c. Perform the necessary calculations to prove that the necessary and sufficient condition of independence is not met.
statistics
. .
== × = 1,2, … ℎ ; = 1,2,… ,

d. Calculate and interpret the Yule's Q coefficient.

PHI ASSOCIATION COEFFICIENT CONTINGENCY TABLE

The phi coefficient or Matthews correlation coefficient as VARIABLE Y

it is also commonly referred to, it allows to calculate the association between
variables measured on nominal, ordinal, or interval scales,
whose data represent authentic dichotomies. For example:
- Yes-No; True-False; Woman-Man; Smokes-Does not smoke, etc.

The asymmetrical association coefficient is defined. of Pearson

like:

Brian W. Matthews, 1938 (Age 77 –

78)Mount Barker, South Australia.

Brian W. Matthews is a biochemist and biophysicist

educated at the University of Adelaide,
PROPERTIES collaborator of the crystallographic methodology
by X-rays at the University of Cambridge and
It is between the values - since 1970 at the University of Oregon as
1 and 1. That is:− ≤ ≤ Physics professor and HHMI researcher at the
If the variables X and Y are independent, Institute of Molecular Biology.
then = He created hundreds of T4 lysozyme mutants (which turned it into the
= if the values on the main diagonal most common structure in the PDB), determined its structure through
if they are zeros (a=d=0), then there exists a
X-ray crystallography and measured their melting temperatures.
complete association between the variables Starting from questions about the basis of mutations "sensitive to
(atributos) X e Y. the temperature", his work has explained much about the effects
energetic and general structural aspects of mutations in the
proteins. Beyond their contributions to biochemistry, Matthews
it is also known in the machine learning community for
the Matthews correlation coefficient, which he introduced in a
paper in 1975. [3] The coefficient is used as a measure of quality
of binary classifications (of two classes).

= − , if the values on the diagonal

secondary (b=c=0) are zeros, then
there is a complete dissociation between the
variables (attributes) X and Y.
Yes > then > , the association is
positive.
Yes < so < the association is
negative.

Interpretation

James Davis offered some expressions that can be used to describe the various ranges of values. Specifically,
Davis developed the expression to use when interpreting Yule's Q, another measure of association. .
EXAMPLE

The question: Should priests marry? Was answered by 90 men and 90 women, as they are classified.
your answers in the following table:
We can ask if there is a relationship between the sex of the subjects and their opinion regarding the celibacy of priests.
catholics.

Yes No
Women 70 20
Men 50 40

Applying the phi formula from Mathews, it is obtained:

70 × 40 − 20 × 50
= = 0.24
√ 90 × 90 × 120 × 60

INTERPRETATION

The phi coefficient, 0.24indicates a weak dependency relationship between sex and opinion regarding celibacy of
priests. Furthermore, since this association is positive = 0.24 > 0that is to say, women are associated with the opinion of
to agree with celibacy and that of men to be in disagreement.
 NATIONAL UNIVERSITY OF SAN AGUSTIN OF AREQUIPA
PROFESSIONAL SCHOOL OF PSYCHOLOGY
COURSE: PSYCHOSTATISTICS PROF.: LICENSEE LUIS GUERRA JORDAN
FIFTH WORKSHOP PART II ASSOCIATION COEFFICIENTS IN TABLES × PHI

In a study to find out if tall men tend to marry tall partners, the following information was published.
about the wives of 134 tall husbands and 116 short husbands. Find the coefficient of association between the
height of wives and husbands.

Tall spouses Low husbands

High handcuffs 112 26
low handcuffs 22 90

Calculate and interpret the phi coefficient or of Mathews.

In a company with 500 workers, with higher education or not, and working in offices or production line.
As shown in the following contingency table. Analyze whether there is an association between the two variables using a
respective measure of association.

Chain of
Offices
production
Studies
270 30
superiors
No studies
80 120
superiors

Calculate and interpret the phi coefficient or of Mathews.

 CHI-SQUARE ASSOCIATION COEFFICIENT OF PEARSON
CONTINGENCY TABLE
VARIABLE Y

ℎ: ú
: ú

An alternative expression to the above for the

the calculation of Chi Square is:

Karl Pearson, 1857-1936(London27 of

Marchof1857- London,27 of
Aprilof1936)

Born in London in 1857 and died in 1936, his

PROPERTIES
family is originally from Yorkshire. Son of a
No expected frequency is less than 1. lawyer, studies at University College
At least 80% of the expected frequencies are School. In 1873, at the age of 16, he was
greater than 5. withdrawn from school for health reasons, and
If these conditions are not met, it cannot be done.
spend the following year with a tutor
apply the test. In such cases we must group the
modalities or increasing the sample size with the In 1875 he obtained a scholarship to King’s College, Cambridge.
objective of ensuring that the conditions are met he said that Cambridge gave him pleasure in friendships, pleasure in
validity of the test.
controversies, pleasure in studying, pleasure in the search for new
For 2 x 2 tables, the most suitable ones are the
lights, both in mathematics and in philosophy and religion;
following conditions:
The marginal frequencies are greater than as well as help to maintain their scientific radicalism within
10
the moderate and reasonable limits. At 22 years old, he moves to Germany
All expected frequencies are greater than 5
If these conditions are not met, it must be applied and studies law, physics, and metaphysics. Between 1880 and 1884, he is a professor
another test known as Fisher's exact test. of mathematics at King College and at University College. In
In the case of independence between the variables1911 was Galton's first professor of Eugenics, the emerging
(attributes) X and Y, the value of = .
the part of Biology responsible for studies aimed at
The higher the value of Chi Square , achieving the improvement of species. He was a convinced Darwinist.
greater will be the degree of association between the
variables X e Y.
To the extent that 2as it approaches zero, the dependence or association will be weak, to the extent that it moves away,
dependency or association will be stronger.
The chi-square coefficient establishes establishes the existence or absence of association between two variables, but not
measure the magnitude of the association.

CHARACTERISTICS

Indicates whether there is a relationship between the variables, but does not indicate the degree or intensity of the relationship as the size increases.
of the sample.
It does not indicate the meaning of the relationship.

It is applicable to variables measured on nominal, ordinal, interval, or ratio scales. Simply, the last three
scales should be reformulated as categorical.
It is applicable when the theoretical (expected) frequencies are not less than five.
Si calculamos el valor de for a contingency table of two rows by two columns, the following holds
equivalence
EXAMPLE

The following table shows the results of a study to find out if drug consumption is very related to
the antisocial behavior of a sample of randomly selected young people in a rehabilitation center.

Since it involves two qualitative variables, at least one is nominal,

the association coefficient (relationship) 'Chi' can be used
square .

For which we calculate the expected frequencies:

1. .1(81)( 88) 1. .2(81)(118)

11= = = 24.41 12 = = = 32.73
292 292
2. .1(93)(88) 1. .3(81)( 86)
21 = Drug consumption
13 =
= = 28.03 = = 23.86
292 292 Total
Behavior Low Medium High
antisocial
3. .1(67)(88) 2. .2(93)(118)
31 = = = 20.19 = = 37.58 Insomnia 25 38 18 81
292 22 = 292
Aggressiveness 20 38 35 93
4. .1(51)88) 2. .3(93)86) Psychotic 18 22 27 67
41 = = = 15.37 23 = = = 27.39
292 292 Normal 25 20 6 51
Total 88 118 86 292
3. .3(67)(86) 4. .3(51)(86)
… 33= = = 19.73 … 43= = = 15.02
292 292

Thus we obtain the contingency table of the expected frequencies as shown below:

Behavior Drug use

antisocial Under Medium High
Insomnia 24.41 32.73 23.86
Aggressiveness 28.03 37.58 27.39
Psychotic 20.19 27.08 19.73
Normal 15.37 20.51 15.02

Then, we calculate the Chi-square coefficient. applying the formula:

Replacing the obtained values:

[25 − 24.41 ]2 [20 − 28.03 ]2 [18-20.19 ]2 [25 − 15.37 ]2 [38 - 32.73 ]2 [38- 37.58 ]2
= + + + + + +
24.41 28.03 20.19 15.37 32.73 37.58

[22 − 27.08 ]2 [20 - 20.61 ]2 [18- 23.86 ]2 [35 −27.39 ]2 [27-19.73 ]2 [6-15.02 ]2
+ + + + + + =
27.08 20.61 23.86 27.39 19.73 15.02

= .

INTERPRETATION

The Chi square coefficient = . > indicates that drug use is related to behavior
antisocial.
 NATIONAL UNIVERSITY OF SAN AGUSTIN OF AREQUIPA
PROFESSIONAL SCHOOL OF PSYCHOLOGY
CURSO: PSICOESTADISTICA PROF.: LIC. LUIS GUERRA JORDAN
SIXTH WORKSHOP PART II ASSOCIATION COEFFICIENTS TABLES × CHI SQUARE FROM PEARSON

From a survey conducted among a group of young people about the current situation, the following results have been obtained:

a. Complete the contingency table by number of people (simple percentage frequency) considering the variable
independiente X (Nivel de estudios) y la variable dependiente Y (Nivel de atención que prestan a las noticias,
issues or events that occur in other countries.

Nivel de estudios Most concerning problems

strike delinquency Total housing
Primary 20 5 5
Secondary 12 7 1
Superior 18 8 4
Total

b. Create grouped bar charts by absolute frequencies (number of young people) and by percentages.
(percentage of youth) considering the level of education on the horizontal axis.

c. To find out if the level of education is related to the problems of the current situation, using the coefficient.
Chi Square. Calculate it and interpret it.
CONTINGENCY COEFFICIENT OF PEARSON

The Contingency Coefficient C or Pearson's C coefficient. Created byKart Pearson(1904), average

the degree (magnitude) of association or relationship between two sets of attributes. It applies to data from
nominal scales and is calculated from frequencies organized in contingency tables of
any number of cells. Calculated from a contingency table will have the same value as
margin of the order of categories in the rows and columns.

The asymmetric association coefficient is defined. as contingency:

=√ ≤ <
+

The upper limit, or maximum value, of the contingency coefficient C depends on the size of the table.

When the table is square, that is, the number of rows it is the same number of columns the maximum value of C can be
calculate using the following equation:

−
á =√ , =

ℎ: ú
: ú

For example,
For contingency tables2 × 2the maximum value is 0.707, that is:0 ≤ C ≤ 0.707
For contingency tables3 × 3the maximum value is0.816, that is to say:0 ≤ C ≤ 0.816
It is only with contingency tables larger than 5 x 5 that the upper limit exceeds 0.900.

The following relationship is fulfilled

−
≤ á = √ <

If the table is not square, the number of rows and columns is not equal. ≠ then the maximum value or upper limit of
C can be calculated using:

− −
á = √ ×

Where h is the number of rows and k is the number of columns.

The following relationship is fulfilled

− −
≤ á = √ × <
STANDARDIZED CONTINGENCY COEFFICIENT C

There is a simple solution to the problem of variation in the upper limits of contingency coefficients. They
they can be normalized or standardized by dividing by their upper limits. This makes all the maximum values,
Regardless of the size or shape of the table, they should be equal to 1. These standardized coefficients can be compared.
between tables of any size, using the following equation:

=
á

This standardized coefficient varies between 0 and 1, that is:

≤ ≤

Classification criteria for the coefficients of C (or they are not very common to be found. Most of the
authors cite only that values close to 0 represent weak or no association, and the strongest association is
For values close to 1, however, the magnitude of these factors is not linear, which interferes with interpretation.
We will follow the following classification ([Link]/Statbook/[Link]:

PROPERTIES

The coefficient C has the following properties:

It is between the values -1 and 1. That is:− ≤ <

When there is a complete lack of association, the coefficient should be null.
When the variables show complete dependence, the coefficient must be equal to one.

According to the definition of the Phi coefficient

A relationship can be established with the Chi square coefficient as:

DISADVANTAGES AND LIMITATIONS

The contingency coefficient C presents some limitations:

The contingency coefficient meets the first characteristic, but does not meet the second, that is, it is equal.
it is zero if there is no association between the two variables or attributes, but it cannot reach one. For that reason
so much
0≤C<1

Two contingency coefficients can only be compared if they come from tables of the same size, because it depends
the upper limit of the number of rows and columns.
C is not directly comparable to any other correlation measure.
The data must be used appropriately for the calculation of the since this test can only be used if at least the
20% of the cells have an expected frequency less than 1.
Its interpretation is difficult because it does not reach its maximum at unity.
EXAMPLE

A researcher aims to find the association between the amount of stress in 167 students of a first-level institute.
semester depending on their socioeconomic conditions.

Frequency and absence of stress.

We first calculate the Chi-square coefficient. of Pearson applying the formula:

Replacing the obtained values:

[0 - 4 ]2 [6− 8.8 ]2 [14 -7.2 ]2 [10 -6 ]2 [16-13.2 ]2 [-68 ]2

= + + + + +
4 8.8 7.2 6 13.2 10.8
= .

The contingency coefficient C is:

Replacing in the formula we have:

.
=√ + .
= .

= .

Next, we calculate the expected maximum value of the contingency coefficient:

−
á = √ = .

Interpretation:

The existing association between stress frequency and socioeconomic condition is moderate, in which as
It is the class of life, significantly decreasing the frequency of the stress process, with a coefficient value.
of contingency of 0.388.
The obtained value of C=0.388 is compared with á = 0.816 is approximately 47.54%. The relationship therefore
is moderate.
 NATIONAL UNIVERSITY OF SAN AGUSTIN DE AREQUIPA
PROFESSIONAL SCHOOL OF PSYCHOLOGY
COURSE: PSYCHO-STATISTICS PROF.: LIC. LUIS GUERRA JORDAN
SEVENTH WORKSHOP PART II ASSOCIATION COEFFICIENTS IN TABLES × DE PEARSON

From a survey conducted with a group of young people about the current situation, the following results have been obtained:

Calculate and interpret Pearson's Contingency Coefficient C.

A school psychologist is testing three reading methods for dyslexic children. Studying the background of these...
children saw that the experience in families with dyslexic siblings could be a cause of the ineffectiveness of the methods.
in order to obtain some evidence, he measured his students on the two variables and obtained the following healing table:

Calculate and interpret Pearson's Contingency Coefficient C.

CRAMER'S V COEFFICIENT

The symmetric association coefficient is defined

The Vde Cramer coefficient is a modified version of the coefficient
from Cramer as:
of Phi association and is used in larger tables than 2 x 2. In this
In this case, there is no upper limit. This coefficient is obtained by adjusting
Phi to the number of rows or to the number of columns of the table, depending =√ ≤ ≤
[ ( ,) −]
of which of the two is smaller, oscillating from zero to one. A value
Where:
large of V and not of the way the variables are related.
: Sample size.
ℎ: Number of rows
: Number of columns

Conventions for describing the magnitude of association in contingency tables (Rea & Parker, p. 203)

PROPERTIES
Harald Cramer (Stockholm, 25 of
Cramer's V coefficient has the
Septemberof1893 - October 5of1985.
following properties:

It is included between the Harald Cramér was a mathematicianSwedishwhat is

values 0 and 1. That is: ≤ ≤ specialized inmathematical statistics. He/She also did
As it is limited, it is the best of the statistical contributions to the distribution of
association measures, for being more theprime numbersytwin cousins. Taught from 1917 to
easy to interpret. 1958 as a teacher in theUniversity of
If there are two rows or two columns, theStockholm(until 1917 as an assistant professor) and was
coefficients φ and V of Cramer are rector of the same from 1950 to 1961.
equals. The chi-squared ranges from 0 to a value that varies according to the number of data points and
The Cramer's V coefficient is the number of cells. Not having a fixed maximum makes it quite difficult to
regardless of the size of the table and,
interpretation. However, a Swiss man, named Harald Cramer (1893–1985),
therefore, it can be used with tables
very interesting in various matters of the world of statistics, was
greater than 2 x 2.
When the variables are totally reasoning mathematically to arrive at the conclusion that the maximum value
independents V=0. what the invention can have is n(m-1), where n is the number of data and m is the
The greater the association, the greater it is. number of values or categories of the variable that has fewer values. The V
the value of the coefficient. it consists of dividing the chi-square by its maximum, so the result is
from 0 (no relationship) to 1 (maximum relationship).

DISADVANTAGES AND LIMITATIONS

EXAMPLE
A school psychologist is testing three reading methods for dyslexic children. Studying the backgrounds of these
children saw that the experience in families with dyslexic siblings could be a cause of the ineffectiveness of the methods.
In order to obtain some evidence, he measured his students on the two variables and obtained the following healing table:

Calculate and interpret Cramer's V coefficient.

We calculate the expected frequencies:

EXPECTED FREQUENCIES
Reading methods
Family background
Method A Method B Method C
Without dyslexic siblings 4 8.8 7.2
With dyslexic siblings 6 13.2 10.8

We calculate the Chi square coefficient applying the formula:

Replacing the obtained values:

[0 - 4 ]2 [6− 8.8 ]2 [14 - 7.2 ]2 [10 −6 ]2 [16- 13.2 ]2 [-68 ]2

= + + + + +
4 8.8 7.2 6 13.2 10.8
= .

We calculate Cramer's V coefficient by applying the formula:

. . .
=√ == √ = √ = .
[ ( ,) −] [ − ]

Interpretation:
It can be concluded that all the coefficients are above half of the range they can take, without reaching the
maximum. It could be said that it results in a moderate-high association.
 NATIONAL UNIVERSITY OF SAN AGUSTIN OF AREQUIPA
PROFESSIONAL SCHOOL OF PSYCHOLOGY
CURSO: PSICOESTADISTICA PROF.: LIC. LUIS GUERRA JORDAN
EIGHTH WORKSHOP PART II ASSOCIATION COEFFICIENTS IN TABLES × V DE CRAMER

EXAMPLE
The number of households that watch a television program, obtained from a survey directed at a sample of households.
Three communities are gathered in the following table.
It is desired to know the degree of association between both factors for these households through Cramér's V coefficient.

a. Create grouped bar charts by absolute frequencies (number of households) and by percentages.
(percentage of households) considering communities on the horizontal axis.

b. Calculate and interpret Cramér's V coefficient.