You are on page 1of 16

ASSIGNMENT 2

Correlation and Regression Analysis

Introduction

This assignment uses data from an online survey conducted for DELL by a UM research team on online
relationships. Ever since DELL started selling PC systems through its website in 1996, online relationships
with its customers have been pivotal to its marketing strategy. Therefore, the company wants to get
insight into the predictors of relationship commitment.
Description dataset

The SPSS data file Assignment2.sav contains 669 respondents from the 2020 DELL Online Relationship
study. The goal of the survey is thus assessing whether familiarity (FAM), situational normality (SN),
structural assurance (SA), and trust (TR) positively influence relationship commitment (RC). The table
below gives an overview of how these variables were questioned in the survey:
MEASUREMENT INSTRUMENTS
FAMILIARITY a
FAM01 I am familiar with DELL through the press.
FAM02 I am familiar with DELL by visiting the DELL website.
FAM03 I am familiar with DELL by purchasing online on the DELL website.
SITUATIONAL NORMALITY a
SN01 The steps required for ordering on the DELL website are typical of other similar websites.
SN02 The information requested from me while ordering on the DELL website is typical of other
similar websites.
SN03 The nature of the interaction with the DELL website is typical of other similar websites.
SN04 The payment process on the DELL website is typical of other similar websites.
STRUCTURAL ASSURANCE a
SA01 I feel safe conducting online transactions with DELL because of its privacy policy.
SA02 I feel safe conducting online transactions with DELL because it provides a toll-free
telephone number for customer contacts.
SA03 I feel safe conducting online transactions with DELL because of its warranty policy.
SA04 I feel safe conducting online transactions with DELL because of its customer support
policy
TRUST a
TR01 Based on my previous experience with DELL I feel that it is honest.
TR02 DELL cares about its customers.
TR03 I can trust DELL as a customer.
TR04 DELL will not behave opportunistically.
RELATIONSHIP COMMITMENT a
RC01 I feel emotionally attached to DELL.
RC02 DELL has great personal meaning for me.

1
RC03 I feel like part of the “DELL family.”
RC04 DELL means a lot to me.
* all measured on an 7-point Likert scale ranging from 1: totally disagree to 7:totally agree.

DEMOGRAPHICS
GENDER Male (1)

Female (2)
AGE years old
EDUCATION (1) Some high school or less
(2) High school graduate
(3) Some College/Technical School
(4) College graduate or higher
INCOME Gross yearly income: $ _____
LENGTH OF RELATIONSHIP years

Questions

1. Calculate a mean score for FAM, SN, SA, RC, and TR. Specifically, for each of the
variables, take the mean of the associated scale items. State whether these resulting
mean scores are categorical or metric variables.

Questions 1

Do the following in spss:


COMPUTE FAM=(FAM01 + FAM02 + FAM03) / 3.
EXECUTE.
COMPUTE SN=(SN01 + SN02 + SN03 + SN04) / 4.
EXECUTE.
COMPUTE SA=(SA01 + SA02 + SA03 + SA04) / 4.
EXECUTE.
COMPUTE TR=(TR01 + TR02 + TR03 + TR04) / 4.
EXECUTE.
COMPUTE RC=(RC01 + RC02 + RC03 + RC04) / 4.
EXECUTE.
Descriptive statistics were used to analyze the average scores of FAM, SN, SA, RC
and TR. Their average scores were 4.1704, 3.5979, 3.7500, 3.4888 and 4.0366
respectively, and the average scores of each variable were obtained as follows. I think

1
the average scores of these results are measurement variables. The resulting mean
scores for FAM, SN, SA, TR, and RC will be metric variables since they are
continuous numerical values.

2. Explore the bivariate correlations among FAM, SN, SA, TR, versus RC. Generally formulate
all hypotheses (H0 and H1 ), conduct the appropriate statistical tests, visually assess the
assumptions of linearity (using a scatterplot – HINT: use graphs>chart
builder>scatterplot>scatterplot matrix) and normality(using histograms and P-P plots),
check for outliers(using boxplots) and interpret the SPSS output with respect to the
hypotheses. What are the implications of these results for the study?

Questions 2

The bivariate correlations between FAM, SN, SA, TR and RC were investigated
For FAM, H0 indicates that FAM and RC are not correlated, while H1 indicates that
FAM and RC are correlated
For SN, H0 is unrelated between SN and RC, and H1 is correlated between SN and
RC
For SA, H0 means that there is no correlation between SA and RC, and H1 means that
there is a correlation between SA and RC
For TR, H0 means there is no correlation between TR and RC and H1 means there is
a correlation between TR and RC

spss was used to test the correlation, and the results were as follows:

1
The results showed that the correlation coefficient between FAM and RC was 0.520,
which passed the test with a significance level of 5%, indicating that there was a
correlation between the two. But the correlation is not strong. According to the
histogram, the hypothesis of normality is satisfied. It can be seen from the boxplot
that the outliers in FAM are not large and will not affect the subsequent regression
analysis. As can be seen from the histogram, the data satisfy the normal distribution.

1
The correlation coefficient between SN and RC is 0.565, but the correlation is not
strong. The test of significance level is 5%, indicating that there is a correlation
between the two. According to the histogram, the hypothesis of normality is satisfied.
It can be seen from the boxplot that the outliers in SN are not large and will not affect
the subsequent regression analysis. As can be seen from the histogram, the data satisfy
the normal distribution.

1
1
The correlation coefficient between SA and RC was 0.337, which passed the test with
a significance level of 5%, indicating that there was a correlation between the two, but
the correlation was not strong. According to the histogram, the hypothesis of
normality is satisfied. As can be seen from the histogram, the data satisfy the normal
distribution. It can be seen from the boxplot that the outliers in SA are not large and
will not affect the subsequent regression analysis.

1
The correlation coefficient between TR and RC was 0.626, which passed the test with
a significance level of 5%, indicating that there was a correlation between the two.
According to the histogram, the hypothesis of normality is satisfied. It can be seen
from the boxplot that the outliers in TR are not large and will not affect the
subsequent regression analysis.

1
1
Scatter plot matrices and histograms can be used to intuitively evaluate linearity and
normality assumptions, respectively, while boxplots can be used to identify outliers.
The results showed that familiarity (FAM), situational normality (SN) were correlated
with structural assurance (SA), trust (TR) and relational commitment (OR).

3. Briefly state if you would conduct the same statistical test as in question 2 to assess the
bivariate relation between education and RC? If not, what other test would you choose
and why? You do not need to report the output.

Questions 3

The same statistical test as in question 2 will not be used to assess the bivariate
relationship between education and RC because education is a nominal variable and
the variables involved in question 2 are all indicator variables, i.e. continuous
quantitative data. For bivariate correlation analysis between nominal and indicator
variables, Chi-square test or Fisher's exact test can be used to evaluate. The two test
methods are non-parametric tests and do not need to satisfy the assumptions such as
normal distribution.

4. Conduct a simple regression with relationship commitment (RC) as dependent


variable and trust (TR) as independent variable:

‒ Formulate the hypotheses (H0 and H1 )

‒ Conduct the appropriate test


‒ Evaluate model fit. Discuss — in your own words--the rationale behind the R² and
the F-statistic.
‒ Assess the assumptions of homoscedasticity (plot standardized predicted values

1
against standardized residuals) and normally distributed errors (look at the
histogram and P-P plot for the residuals).
- Interpret the SPSS output on the individual predictor. What happens to RC if TR
increases with 1 scale point
-Visualize the relation in a scatter plot. Do you think that there are extreme scores
might be affecting your results (HINT: use “casewise diagnostics”).
What are the implications of these results for the study?

Questions 4

H0: Trust (TR) has no significant effect on relationship commitment (RC)

H1: Trust (TR) has a significant effect on relationship commitment (RC)

RC i=b0 +b 1 TRi +ε i

We use spss software for simple regression:

According to the statistical results, the value of R Square is 39.2%, which means that
trust (TR) accounts for 39.2% of relationship commitment (RC), leaving 60.8% of
relationship commitment (RC) unexplained.

The p-value corresponding to the F statistic is less than 0.1%. We can reject H0 and
conclude that the change in relational commitment (RC) predicted by this model is
significant. According to mean square error and normal distribution error, we can see
that the model fits the relationship between x and y well. According to the histogram,
we can see that the data satisfies the assumption of normal distribution.

1
According to the coefficient test of the fitted model, it is found that the coefficient of
TR passes the significance test, so we can write the final fitted model
RC i=0.768 +0 .674 TRi
When TR= 1,0.768 +0.674=1.442
Therefore, when TR changes by one unit and RC changes by 1.442 units, the results
show that there is a significant positive correlation between trust and relationship

1
commitment, indicating that relationship commitment increases with the increase of
trust.

5. The principal researcher of the team suggests that familiarity (FAM), situational normality
(SN) and structural assurances (SA) might be added as independent variables to the model in
question 4. Is the explanatory power of the model significantly increased by adding
familiarity (FAM), situational normality (SN) and structural assurances (SA) to the model?
Furthermore, please do the following:
‒ Formulate the hypotheses (H0 and H1 )

‒ Conduct the appropriate test

‒ Report the descriptive statistics


‒ Evaluate model fit.
‒ Assess the assumptions of homoscedasticity (plot standardized predicted values
against standardized residuals), normally distributed errors (look at the histogram
and P-P plot for the residuals), and linearity (retrieve all partial plots).
‒ Interpret the SPSS output on the individual predictors.

What are the implications of these results for the study?

Questions 5

H0: Familiarity (FAM), situational normality (SN), structural assurance (SA) and
trust (TR) had no significant influence on relationship commitment (RC)

H1: Familiarity (FAM), situational normality (SN), structural assurance (SA) and
trust (TR) had significant influence on relationship commitment (RC)

RC i=b0 +b 1 TRi +b 2 FAM i +b 3 SN i +b 4 SA i+ ε i

We use spss software for simple regression:

According to the statistical results, the value of R Square was 57.7%, indicating that
the addition of familiarity (FAM), situational normality (SN) and structural assurance
(SA) to the model significantly improved the explanatory ability of the model.

1
The p-value corresponding to the F statistic is less than 0.1%. We can reject H0 and
conclude that the change in relational commitment (RC) predicted by this model is
significant. According to mean square error and normal distribution error, it can be
seen that the model fits the relationship between independent variable and dependent
variable well.

1
According to the coefficient test of the fitted model, it is found that the p values of the
coefficients of TR, FAM, SN and SA are all approximately 0, so they all pass the
significance test. We can write the final fitted equation:
RC i=−0.654+ 0 .454 TRi +0 .171 FAM i+ 0.455 SN i −0.011 SAi
This indicates that there is a significant positive correlation between trust (TR) and
relationship commitment, indicating that relationship commitment will increase by
0.454 units as trust (TR) increases by 1 unit, other conditions being constant.
There was a significant positive correlation between familiarity (FAM) and
relationship commitment, indicating that an increase of 1 unit of familiarity (FAM)
would lead to an increase of 0.171 units of relationship commitment, other conditions
being constant.
There is a significant positive correlation between situational normality (SN) and
relationship commitment. Other conditions remain unchanged, indicating that with the
increase of one unit of situational normality (SN), relationship commitment will also
increase by 0.455 units.
There is a significant negative correlation between structural guarantee (SA) and
relationship commitment. Other conditions remain unchanged, indicating that with the
increase of structural guarantee (SA) by 1 unit, relationship commitment will decrease
by 0.011 unit.

1
6. Could multicollinearity potentially affect the results in question 5? Explain the concept of
multicollinearity for OLS regression analysis, how it potentially could affect your
findings and how it can be diagnosed?

Questions 6

Multicollinearity is a statistical problem that occurs when the independent variables in


a regression model are highly correlated with each other. This can lead to inaccurate
estimates of regression coefficients and reduce the predictive power of the model.
In problem 5, multicollinearity is possible if FAM, SN and SA variables are highly
correlated. We can diagnose by examining the correlation matrix or by calculating the
VIF value of each variable.

7. Summarize the main findings of the above analysis for the senior management of DELL
and provide your recommendations.

Questions 7

It is found that trust, structural assurance, familiarity and situational normality are
positively correlated with relationship commitment. Adding familiarity, situational
normality and structural assurance to the model significantly improves the
explanatory power of the model. If the FAM, SN, and SA variables are highly
correlated, multicollinearity may affect the results. It is therefore recommended that
Dell focus on building trust and providing structural assurance to increase customer
commitment to the relationship. In addition, familiarity and situational normality
should be considered when developing marketing strategies.

You might also like