PSYC 2523/3423 Research Skills 4 Overview

and Correlation Statistical Techniques

Dr Carlo Campagnoli & Dr Gina Koutsopoulou

 You’re invited to participate in our study
Follow the QR code to the Qualtrics questionnaire

• Takes only 5-10 minutes

• Completely anonymous
Overview of
Research Learning Objectives

Skills 4
The module has two related components:

1) Two practicals
Hand-in dates
• Practical 1: Tues 7 March 2023 (noon)
• Practical 2: Fri 5 May 2023 (noon)

2) 10 lectures on research methods & statistics.

 Lectures
1 Advanced Correlation Techniques 6 Introduction to Psychometric
(relates to the analysis of Testing: Questionnaire Design
practical data)
7 Item analysis in questionnaire
2 Multiple regression I design

3 Multiple regression II
8 Factor analysis
4 Preparing for your
Major Project
9 More advanced techniques:
power analysis and meta-analysis
5 Samples and Populations
The British Psychological Society

The BPS “has national responsibility for the

development, promotion and application of
psychology for the public good, and
promotes the efficiency and usefulness of
its members by maintaining a high standard
of professional education and knowledge.”
“The Society was granted its Royal Charter
in 1965.”
• http://www.bps.org.uk
• The BPS Code of Ethics and Conduct
• Code of Human Research Ethics
 Overview of Practical 1

Session 1 Introduction to the Practical

Session 2 Ethics and Data Collection
In this session we will discuss the ethical considerations related to this study and
collect data.
Session 3 Data Handling
In this session students will be shown how to check and handle data in the data file
to produce the appropriate predictor variables
Session 4 Data Analysis
In this session we will discuss your data analyses (Multiple Regression).
Session 5 Writing your research report and Support
Ways for us to get connected
There is support available and several ways for us to get connected
and get in touch and engage in communications:
• Discussion Boards
• Live sessions
• Additional Friday Q&A’s (in case you couldn’t attend Tue/Thu)
Coming to Leeds to ‘read’ Psychology &
Preparing for your selection of a Major
Project

• No magic solution - Reading can unlock your academic

research skills and develop your academic writing

• PSYC3520 Major Project 30 credits

Practicals, Report Writing, Reading
• American Psychological Association (2020).
Publication Manual of the American Psychological Association. 7th Edition.
• Field, A. (2018). Discovering Statistics using SPSS. London: Sage.
A Companion website
• Howitt, D. and Cramer. D. (2017).
Understanding Statistics in Psychology with SPSS. Harlow: Prentice Hall.
A companion website
• Howitt, D. and Cramer, D. (2017). Introduction to SPSS in Psychology.
Pearson/Prentice Hall. A companion website
• Howitt, D. and Cramer, D. (2017). Research Methods in Psychology. Harlow:
Prentice Hall. A companion website
Open the RS4 Module
Overview from Module
information in Minerva and
explore the folder
Aims and Learning Objectives of this lecture

Pointers to knowledge covered in previous Research Skills modules (RS2 in particular)

Researchers are often interested in exploring whether two variables (X and Y)

are associated (or correlated).
• When to use correlational analysis (Research Designs and
Methodologies)
• How to statistically test for associations (using SPSS)
• using: Pearson’s, point biserial rpb , Spearman, Kendall Tau-b; φ2
crosstabs/contingency tables; and partial, semi-partial correlations
• Interpret Scatterplots, the correlation coefficients and the coefficient of
determination r2
• Introduce other ways to use correlations
Aims and Learning Objectives of this lecture

Pointers to knowledge covered in previous Research Skills modules (RS2 in particular)

Researchers are often interested in exploring whether two variables (X and Y)

are associated (or correlated).
• When to use correlational analysis (Research Designs and
Methodologies)
• How to statistically test for associations (using SPSS)
• using: Pearson’s, point biserial rpb , Spearman, Kendall Tau-b; φ2
crosstabs/contingency tables; and partial, semi-partial correlations
• Interpret Scatterplots, the correlation coefficients and the coefficient of
determination r2
• Introduce other ways to use correlations
 Variables (X, Y)

• Examples (aggression, reaction time, weight)

The • Measurement (scores, in seconds, kilos)

Predictable (Ho and H1,

Science of 1- or 2- tailed) relationships
Psycholog • Causal or Correlational
• A change in one variable may produce or be associated

y with a change in another

• Independent (X) and Dependent (Y) variables (IVs and
DVs)

How do we study these statistical relationships?

 Correlation (association) versus Causation
Probability
that a train
is coming 1.2

“X and Y are correlated”

1

0.8
r = .93

0.6

0.4

0.2

0
0 20 40 60 80 100 120

# people waiting for the train on platform

 Correlation (association) versus Causation
Probability
that a train
is coming 1.2

“X predicts Y”
1

0.8
Y=a+b*X

0.6

0.4

0.2

0
0 20 40 60 80 100 120
What if we tested “X causes Y” empirically?
# people waiting for the train on platform
Correlation (association) versus Causation

Neither correlation not regression imply causation by default, they only

describe associations between variables

Causation is proven through logic arguments and supported by evidence

(hypothesis testing)
Ways to collect data depending
on the Research Question

Source

Experimental design Survey design (correlational)

Systematic manipulation of one or more Quantitative description of trends, attitudes, or
variables (X) to evaluate an outcome (Y) opinions of a population
Creswell (2018)
Holds other Research
variables constantDesign
to isolate effects Testing association of X and Y
 Case Gender Age IQ (Y) Self-Esteem (X)

Correlation 1 F 18 135 8

al Research 2 M 19 120 4

Design 3 M 25 100 8

4 F 21 105 3

The basic correlational design is one in which the researcher measures two (X,Y) or more (Z)
different variables at the same time using a single group of cases/participants
These variables might be age, IQ (Y), and self-esteem (X)
It investigates whether or not there is a correlation (a statistical association) between the
chosen variables.
n of the relationship between 2 variables X and Y

Positive correlation between two variables Negative correlation between two variables
Near-zero correlation between two Influence of outliers on a correlation
variables
 What is a correlation coefficient?

It is the result of a
statistical test measuring
It is a standardised
the extent to which two
statistical index describing
variables X and Y are
statistically related

How much two variables vary together Direction Effect Size

How much two variables vary on their own
 +1.0 +0.9 +0.8 +0.7 +0.6 +0.5 +0.4 +0.3 +0.2 +0.1

0.0

-0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 -0.9 -1.0

the correlation-coefficient r varies between -1 and +1

0 = no relationship
Types of Correlation Coefficients
Categorical

Correlation X,Y

Continuous Discrete data

Parametric Ranked Point Biserial correlation

Pearson’s r Few tied ranks Tied Ranks

Spearman’s Rho Kendall’s Tau τ

Two binary φ
 Basic Questions
Which association test? It depends on the variables’ scales
Variable Composition Appropriate Tests
Two numerical variables Correlation coefficient (e.g.
Pearson, Spearman or
Regression)
Two categorical Contingency tables (e.g. Chi
variables Square, Fisher or McNemar
tests)
One categorical IV and T-test (two levels) or ANOVA
one nominal DV (three or more levels)
Andy Field

IF there were a relationship between these 2 variables X and Y, then as one variable deviates
from its mean, the other variable should deviate from its mean in the same or opposite way
(from Field 8.2) The more they covary and change in a similar way the more covariance
there will be.

In RS2 we used the Pearson r (or any test) by applying -

by hand- the 4-Step Hypothesis Testing Technique
 Andy Field

Participant 1 2 3 4 5 Mean SD
Advertisements watched (X) 5 4 4 6 8 5.4 1.67
Toffee packets bought (Y) 8 9 10 13 15 11.0 2.92

Mean of Y = 11

Mean of X = 5.4

Graphical display of the differences between the observed data

and the means of two variables
Andy Field

How to report correlation coefficients

X (adverts watched) and Y (packets bought) do not significantly

correlate (r(3) = .87, p = .06) or

There was no significant statistical relationship between the number of adverts

watched and the number of packets purchased (r(3) = .87, p =.06).
Aims and Learning Objectives of this lecture

Pointers to knowledge covered in previous Research Skills modules (RS2 in particular)

Researchers are often interested in exploring whether two variables (X and Y)

are associated (or correlated).
• When to use correlational analysis (Research Designs and
Methodologies)
• How to statistically test for associations (using SPSS)
• using: Pearson’s, point biserial rpb , Spearman, Kendall Tau-b; φ2
crosstabs/contingency tables; and partial, semi-partial correlations
• Interpret Scatterplots, the correlation coefficients and the coefficient of
determination r2
• Introduce other ways to use correlations
Pearson’s r - Scores on musical and mathematical ability
for 10 children(Howitt and Cramer, 2014)
Open SPSS and create the data file
Scatterplot – how to 3

2
Example of a scatterplot and line of best fit
Correlation – how to
3

1
Interpretation of Pearson’s r (SPSS Output)

The correlation coefficient is −.90.

The relationship is negative and statistically significant .

In a report, we would write:

‘There is a significant negative relationship between musical ability and mathematical ability, r(8) = −.90, p < 0.001.
Children’s musical ability scores increase mathematical ability scores decrease’
Effect size measures for correlational designs

r = .1 (small effect)

r = .3 (medium effect)

r = .5 (large effect)
Coefficient of determination R2

By squaring the value of r we get the proportion of

R
variance in one variable shared by the other.
2 For example a coefficient of r = 0.6 indicates that
36% of the variance of X and Y is shared

R2 = (0.6 × 0.6) × 100 = 36%

Coefficient of determination R2
Table 8.4
Variance shared by two variables
Coefficient of determination R2
By squaring the value of r you get the proportion of variance in one variable shared by
the other.

Y
Y
X

0.1 = 1%
0.3 = 9% X
0.5 = 25%
0.9 = 81%
More overlap = More shared variance
 Use when one of the variables is a dichotomous variable
(categorical: nominal with only two categories)
Point-Biserial • Is someone a student?
• Treatment / control
correlation: and the other variable is continuous.

Analysis and
Interpretation Calculation is simple – Same as Pearson’s in SPSS
Example: The relationship between gender (X) and
estimation of own IQ (Y)
code gender (1=male, 2=male) own IQ

Carry out a Pearson correlation with SPSS

Point-Biserial correlation
“IQ_Practical_Real_DATA.sav”
Point-Biserial correlation
“IQ_Practical_Real_DATA.sav”
Grouping variable (IV) Estimated own IQ score (DV)
1 = Male, 2 = Female
1 220
1 225
1 205
1 210
1 250
1 220
1 215
1 211
2 101
2 131
2 103
2 100
2 120
2 130
2 100
2 100
Point-Biserial correlation – how to
Calculation in SPSS is simple – Same as Pearson’s
• The relationship between estimating own IQ (Y)
• and gender (X)

IV  Code: Gender (1=male, 2=female)

DV  Own IQ (estimation of own IQ)
Point-Biserial correlation – how to
H1: There is a relationship between gender (X) and estimating own IQ (Y)

Mean for Female = 100.97, SD=18.47

• What about the sign of the
correlation?
• Ignore the sign, it depends
on the coding.
– Check the means.

Mean for Male = 113.10, SD=17.89

Comparative design: t = 91.34, N = 281, p < 0.01

More examples
 Interpretation of r
Exam Anxiety.sav

This box shows that the

correlation between X and Y is r =
0.397.

The 2-tailed significance or

probability level is p =.001 or less
so the correlation is statistically
significant.

The number of cases on which this

correlation is based is 103
In SPSS a "correlation between the two variables (name the (therefore df= 101).
variables) is significant as p < .001"
 EXERCISE
How much variance do exam anxiety (X) and exam performance (Y) share?
Nonparametric correlations 1/2

Spearman’s rho (rank correlation coefficient)

• Pearson’s correlation test on the ranked data

Kendall’s tau-b

• Better than Spearman’s for small samples

World’s biggest liar competition

• See from Field (8.4.3 & 8.4.4)  The Biggest Liar.sav

Nonparametric correlations 1/2
Nonparametric correlations 2/2
The Phi (φ) Correlation (2x2)
 Nonparametric correlations 2/2

2x2 design
Categorical data
Frequencies: How many?

Testing with a version of the statistical test

χ2 ; the Phi (φ) Correlation
Nonparametric correlations 2/2

SPSS commands: Graphs, Legacy

Dialogs,
•Bar Clustered, Summary for Groups
of
•Cases, Category is gender
•Define clusters by Careers
 Partial and Semi-Partial Correlations

(c) Partial correlation:

• Measures the relationship between two variables, controlling for the

effect that a third variable has on them both.
• Useful when determining whether the relationship between two
variables (e.g., exam performance and anxiety) is explained by a third
variable (e.g., hours of study).

(d) Semi-partial correlation (in Multiple Regression):

• Measures the relationship between two variables controlling for the

effect that a third variable has on only one of the variables in the
correlation.
• Useful when trying to explain the variance in one particular variable for
a set of predictor variables.

Murray and Conner, 2009

Partial Correlation Output and Interpretation

The top part of this table gives

the zero-order correlations
when there are no (-none-a)
Control Variables.
The zero-order correlation
between performance (Y) and
anxiety (X) is r = -.441

The bottom part gives the

partial correlations. The
numerical correlation between
performance and anxiety
partialling out time spent
revising is r = -.247.
Aims and Learning Objectives of this lecture

Pointers to knowledge covered in previous Research Skills modules (RS2 in particular)

Researchers are often interested in exploring whether two variables (X and Y)

are associated (or correlated).
• When to use correlational analysis (Research Designs and
Methodologies)
• How to statistically test for associations (using SPSS)
• using: Pearson’s, point biserial rpb , Spearman, Kendall Tau-b; φ2
crosstabs/contingency tables; and partial, semi-partial correlations
• Interpret Scatterplots, the correlation coefficients and the coefficient of
determination r2
• Introduce other ways to use correlations
Other ways to use correlations
Psychometrics and Questionnaire Design (https://ipip.ori.org/)

As researchers in psychology, we measure constructs (psychological variables) through

questionnaires/scales and the items in these measurement tools operationalise a definition of
such constructs
Psychological attribute Construct Operationalisation (definition) Measurement tool (eg questionnaire)
Other ways to use correlations
Psychometrics and Questionnaire Design (https://ipip.ori.org/)

Standardization: it’s important that our measurement tool is tested for its:

• Reliability
IQ IQ
• Validity
Unreliable Reliable

IQ

Reliable
+
Valid
Exercises
Survey.sav
As Perceived
stress decreases,
so does the
amount of
control over
internal states
Interpretation:
Is the r statistically
significant?

What is the effect

size?

As Perceived stress
decreases, so does the
amount of control
over internal states
What statistical analysis? childaggression.sav

• Parenting style (high=bad),

• computer games (high = more time),
• television, diet (high = good diet low in E) and
• sibling aggression (high = more aggression seen in sibling).
Conclusion
When exploring the association between two variables it is
important to first inspect the scatter plot
The Pearson correlation coefficient (or other) is a statistical
indicator of the relationship
They inform you about the strength and direction of the
relationship as well as the amount of variance shared
The correlation coefficient is itself a measure of effect size
(Cohen’s 1988).