Correlations

Madgerie Jameson and Vivian Alexander

02/18/2025 1
UWI School of Education
 Objectives
At the end of this lecture students will be
able to:
1. Define correlations and explain how they
work.
2. Compute correlation coefficient
3. Interpret the values of the correlation
coefficient
4. Interpret the coefficient of determination.

02/18/2025 Madgerie Jameson UWI School of Education 2

 What are Correlations?
A correlation is a measure of
association between or among
variables.
•expressed as a numerical index
called the correlation coefficient.

02/18/2025 Madgerie Jameson UWI School of Education 3

 Correlation Coefficient
The correlation coefficient ranges
between -1 and +1.

They can be either

•Positive correlation if the variables
changes in the same direction.
• Negative correlation if the
variables change in opposite
directions.
02/18/2025 Madgerie Jameson UWI School of Education 4
 Corresponding relationship between
variables

Change in Change in Correlation Value Example

Variable x Variable y

X increases in Y increases in The more time you spend studying the

value value Positive .00 to + 1.00 higher your test scores will be.

X decreases in Y decreases in Positive .00 to +1.00 The less money you put in the bank the
value value less interest you will earn.

X increases in Y decreases in
value value Negative -1.00 to .00 The more you exercise the less you weigh.

X decreases in Y increases in The less time you take to complete a test

value value Negative -1.00 to .00 the more errors you will make.

02/18/2025 Madgerie Jameson UWI School of Education 5

 Correlational Research
•Correlational research seeks to
discover relationships among
variables.

•Relationship means that an

individual’s status on one variable
tends to reflect his or her status
on the other.
 Correlational Research

CAUTION:
CAUTION:In Incorrelational
correlationalresearch
researchyou
youcan
canonly
onlyestablish
establish
that
thatthe
thevariables
variablesare
areRELATED
RELATEDor orASSOCIATED
ASSOCIATEDwith with
each
eachother.
other.
You
YouCAN
CANNOT NOTsaysayone
onevariable
variablecauses
causeschanges
changesin
inthe
theother.
other.
This
Thiscan
canonly
onlybebedone
donethrough
throughexperimental
experimentalresearch.
research.
 Which correlation to use?

Phi
correlation
Pearson
Spearman Product
rho Moment

Biserial
Correlation
Kendall
tau
 Types of Correlation
Measurement Level Measurement Level
Correlation
Variable 1 Variable 2
Interval (Continuous) Interval (Continuous) Pearson

Real Dichotomy Interval (Continuous) Point Biserial

Artificial Dichotomy Interval (Continuous) Biserial

Real Dichotomy Real Dichotomy Phi

Artificial Dichotomy Artificial Dichotomy Tetrachoric

Ranking Ranking Spearman rho

Ranking Ranking Kendall’s tau

 Pearson Product-Moment
Correlation
•The Pearson-product moment (r)
correlation is the most famous
and widely used correlation.
•measures the linear association
between two variables.
•r = 0 does not mean uncorrelated
variables, but rather a possibly non-
linear relationship.
 Pearson Product-Moment
Correlation
•Used when both variables are
continuous such as test scores,
experience, money, etc.

•It is sensitive outliers or extreme

variables
Spearman (rho) and Kendall (tau)
•Spearman-rank correlation (rho),
is another very famous and widely
used correlation.

•Created as an alternative to the

Pearson correlation.
Spearman (rho) and Kendall (tau)
•Robust to outliers

•Used when both variables are

rank ordered.
•More than 20 cases – Spearman rho
•Less than 20 cases – Kendall tau
 Point Biserial Correlation
•When the predictor variable is a natural
(real) dichotomy (two categories) and
the criterion variable is interval or
continuous, the point biserial
correlation is used.

•E.g., the correlation between Gender

(dichotomous) and Height (continuous).

•This correlation is very common in

“Item-Analysis” studies
 Biserial Correlation
•When the predictor variable is an
artificial dichotomy (two
categories) and the criterion
variable is interval or continuous ,
the biserial correlation is used.

•E.g., Correlation between height

(tall or short) and arm length
 Other Correlations
• You can perform multiple correlations using such
approaches as partial correlation, multiple
regression, discriminant analysis, and factor
analysis.
• These are outside the scope of this class.
 Principles to Remember
•For correlations, there must be at
least two measures.

•A correlation may be statistically

significant but be weak or low
which means it is nothing to get
excited about. It has no practical
significance.
 Correlation Strength
•The absolute value of the
correlation reflects the strength of
the correlation. So a correlation
of -.70 is stronger than a
correlation of +.50.

•A correlation always reflects the

situation where there are at least
two data points or variables per
case.
 Correlation Strength
Size of Correlation Coefficient General Interpretation

.8 to 1 Very strong relationship

.6 to .8 Strong relationship

.4 to .6 Moderate relationship

.2 to .4 Weak relationship

.o to .2 Weak or no relationship

This is effective for a quick assessment of the strength of a relationship.

A more precise way to interpret correlation coefficient is by computing
the coefficient of determination
• Do not assign a value judgment to the sign of the
correlation, e.g. + being good and – being bad.
correlation.

• Pearson product-moment correlation is

represented by the small letter r with a subscript
representing the variables that are being
correlated.
• rxy is the correlation between variable x and variable
y
• rweight-height is the correlation between weight and
height.
• rSAT*GPA is the correlation between SAT score and
grade point average ( GPA).
02/18/2025 Madgerie Jameson UWI School of Education 20
 Correlations
•Scatter plots are often used to
depict correlations
6000

5000
Calories per day

4000 This
Thischart
chart
shows
showsaa
3000
strong
strong
2000 positive
positive
correlation
correlation
1000
0
100 150 200 250 300 350 400
Weight
Correlation/Scatterplot
Correlation/Scatterplot
 Limitations of the
correlation coefficient
• Correlation does not mean causation.

• The Pearson’s r correlation coefficient

assumes linear relationships, that is
relationships where higher scores on x are
linearly related to higher scores on y. Not
all relationships follow this form. There are
curvilinear relationships, for example,
 Limitations continued

• Correlations are affected by the range of

the variables
• If the variable range is small the
correlation coefficient will be artificially
low.
• This is called range restriction (restriction
of range).

•Correlations can be affected by outliers

(unusual cases)
02/18/2025 Madgerie Jameson UWI School of Education 25
 The Coefficient of
Determination
• The percentage of variance in one variable
that is accounted for by the variance in
the other variable.

• To determine exactly how much of the

variance in one variable can be
accounted for by the variance in another
variable, the coefficient of determination
is computed by squaring the correlation
coefficient.
 For Example

•The correlation between two

scores .70
•The coefficient of determination =
r2 = .72 or .49.

• This means that 49% of the

variance in your test results can
be explained by the variance in
your study time.
 However
• If 49% of the variance can be explained
this means that 51% cannot. Even in a
strong correlation of .7.

• The amount of unexplained variance is

called the coefficient of alienation or
coefficient of non-determintation.
 How variables share variance
and the resulting correlation
correlation Coefficient of Variable X Variable Y
determination

Rxy = 0 R2xy = 0 X 0% Y
shared
Rxy = .5 R2xy = .25
or 25% x y 25%
shared

Rxy = .9 R2xy =.81 81%

or 81% xy shared

29
 Summary
•Correlation describes the
association among variables.

•Correlations - 1.0 to + 1.0.

•The coefficient of determination( r2)

tells us the present variance in the
dependent variable (Y) that is
explained by the independent
variable (x).