Scribd
Upload
23 views

Steps in SPSS To Find Correlation Matrix and Partial Correlation

The document discusses correlation and partial correlation analysis techniques. It provides details on Pearson's product-moment correlation and how to calculate it using SPSS. It also discusses assessing normality and linear relationships between variables. The document then explains how to perform a partial correlation to measure the relationship between two variables while controlling for a third variable.

Uploaded by

manas.23pgdm157
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
23 views

Steps in SPSS To Find Correlation Matrix and Partial Correlation

The document discusses correlation and partial correlation analysis techniques. It provides details on Pearson's product-moment correlation and how to calculate it using SPSS. It also discusses assessing normality and linear relationships between variables. The document then explains how to perform a partial correlation to measure the relationship between two variables while controlling for a third variable.

Uploaded by

manas.23pgdm157
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 33

Correlation and Partial

Correlation
• The association between variables is better explained by strength,
direction, shape and other dimensions as per the requirement of the
analysis.
• The technique used in research to investigate the relationship of one
variable to another is called bivariate correlation analysis.
• This correlation coefficient is denoted by r, and it measures the
association between two variables in terms of strength (strong or weak)
and direction (positive or negative).
• The correlation coefficient (r) ranges from −1.0 to +1.0. The value closer
to 1 indicates strong association, and closer to + or – sign indicates
positive and negative association, respectively.
Correlation Coefficient

Range of r Interpretation Range of r Interpretation


Positive Negative
0.1 to 0.4 Weak positive 0.1 to 0.4 Weak negative
0.5 to 0.6 Moderate 0.5 to 0.6 Moderate
positive negative
0.7 to 1.0 Strong positive 0.7 to 1.0 Strong
negative
Partial Correlation
• The decision of variables among which partial correlation needs to be
computed depends upon objective of the study.

• In computing partial correlation, one of the variables is usually a


criterion variable, and the other is the independent variable having
the highest magnitude of correlation with the criterion variable.

• Criterion variable is the one in which variation is studied as result of


variation in other independent variables.
• Usually criterion variable is known as dependent variable.

• Here the criterion variable is the Customer loyalty because the effect
of other variables need to be investigated on it.
Research Problem and Test Technique

• Data file multiplex.sav presents the data of randomly selected 60


households depicting the frequency of visits to multiplex by families
on annual basis along with their monthly income in US$. Both these
variables were measured at ratio scale.
Framing Hypothesis

Null hypothesis H0 : ρ = 0
(Population correlation coefficients (ρ) is equal to zero.)

We can also express null hypothesis indicating no relationship


between the variables.
H0: No significant correlation exists
between the two variables.

Alternative hypothesis Ha : ρ ≠ 0
(Population correlation coefficients (ρ) is not equal to zero.)
Executing Pearson’s Product-Moment
Correlation with SPSS Procedure
• Assessing Linear Relationship
Exhibit 12.1. Use multiplex.sav » Graphs » Legacy Dialog
» Scatter plot » Select visit_multiplex and transfer in Y
Axis » Select Income transfer in X Axis » Click OK
Result of Linear Relationship

• The scatter plot indicates a


pattern of linear
relationship.
• It can be observed that as
income tends to increase,
the frequency of visits also
increases, thus showing a
positive relationship
between the variables.
• Assessing Normality
Exhibit 12.2. Use multiplex.sav » Menu bar » analyse
» Non-parametric test » Legacy Dialogs » One sample
KS test » Transfer visit and Income to Test Variable(s)
List » Click on Normal Distribution under Test
Distribution » Click OK
Results of Normality

Table 12.2. Assessing Normality with One-Sample KS

N Normal Parameters K-S Z


Sig. (2-
Mean Std. Deviation tailed)

Visits 60 6.5 3.93 0.102 0.190


Income 1213.33 610.21 0.105 0.163
The p-values for visit (p > 0.05, 0.190) and income (p >
0.05, 0.163) are more than 0.05; hence, we fail to reject
the null hypothesis of normality and conclude that both
the variables are normal distributed.

Lokesh Jasrai (2020), SAGE TEXTS


• SPSS Path for Descriptive Statistics and Correlation
Coefficient
Exhibit 12.3. Use multiplex.sav » Menu bar » analyse »
Correlate » Bivariate » Select and transfer visit and Income to
Variables box » Click Options and select Mean and standard
deviation (optional if interested) under Statistics » Click OK
Partial Correlation

• The partial correlation measures the strength and


direction of a linear association between two variables
after controlling the effects of one or more other
independent variables in the study (Brown and
Hendrix 2014).
• Simplest partial correlation includes three variables,
one dependent and independent variable and one
control variable. The result of partial correlation is also
presented as correlation coefficient (rpartial).
• The value of rpartial ranges from −1.0 to +1.0 (Salkind,
2010). The subscript indicates correlated variables
along with the controlling variable.
Research Problem and Test Technique

• In the partial correlation, investigating the


relationship between frequencies of visits and
monthly income, we also want to ascertain the
effect of family size as covariate in the initially
correlated variables.

• In view of this, we use all three variables in the


data set multiplex.sav to examine the relationship
between visits and monthly income after
controlling the family size.
Framing Hypothesis
Null hypothesis
We can express the null hypothesis, indicating no relationship
between the variables.

H0: (When removing the effect of control variable, no significant


correlation exists between the two variables.)

Alternative hypothesis Ha: ρ ≠ 0


(When removing the effect of control variable, population correlation
coefficients (ρ) is not equal to zero.)

Lokesh Jasrai (2020), SAGE TEXTS


Executing Partial Correlation with SPSS
Procedure
Assessing Linear Relationship
Exhibit 12.4. Use multiplex.sav » Graphs » Legacy Dialog »
Scatter plot » Matrix scatter » Select visits, Income and family
size (pressing shift) transfer in Matrix Variables box » Click OK
Result: Matrix Scatter Plots

• A linear positive
association can be seen
between visit-income,
whereas negative
associations appear for
the pairs visit-family size
and income-family size.
Reporting the Results of Pearson’s Product-
Moment and Partial Correlation
• A strong positive Pearson’s product-moment
correlation exists (r = 0.638; p < 0.05, 0.000) between
visit (X) to multiplex and family household income (Y).

• A strong negative correlation (r = −0.675; p < 0.05,


0.000) exists between visit (X) to multiplex and family
size (Z) and between household income (Y) and family
size (Z).

• Finally, the partial correlation between visit (X) to


multiplex and family household income (Y) is significant
but week positive (r = 0.358; p < 0.05, 0.000) in nature.
Correlation Output
Partial Correlation Output
Spearman Correlation
• If one of the data is categorical

• Use YouTube Survey.sav file >> Menu bar >> Analyze >> Correlate >>
Bivariate >> select ‘minutewatch’ and ‘educ’ in variables box >> Select
spearman
• Kendall’s Tau is more effective for small sample size.

• Here Spearman’s correlation is better, as sample size is large.

You might also like