CHAPTER 4: STATISTICAL REPORT

Analyzing the Relationship Between Two Quantitative Variables

1. INTRODUCTION

1.1 Research Question

The primary research question for this study is: Does the reputation of a seller or Key Opinion Leader
(KOL) significantly affect a consumer's purchase intention?

1.2 Variables

• Response Variable (Y): Purchase Intention (measured on a Likert scale from 1 to 5). This is the
dependent variable as it represents the outcome being predicted.
• Explanatory Variable (X): KOL/Seller Reputation (measured on a Likert scale from 1 to 5). This is
the independent variable because a seller's reputation is an established factor that precedes and
theoretically influences the consumer's buying decision.
giải thích lý do lựa chọn 2 biến
1.3 Data Collection

Data was collected via an online survey directly from respondents. The sample size is n = 33
observations. The data relies on self-reported measurements evaluating both variables concurrently.
thời gian khảo sát
cách thu thập dữ liệu
2. DESCRIPTIVE STATISTICS

2.1 Descriptive Summary Table

Statistic KOL Reputation (X) Purchase Intention (Y)

Mean 3.3333 2.8788

Median 3.0000 3.0000

Standard Deviation 1.3617 1.3638

Minimum 1 1

Maximum 5 5

Count (n) 33 33
2.2 Comments

Based on the descriptive statistics, both variables have similar ranges (1 to 5). The standard deviation
for KOL Reputation (X) is 1.3617, while for Purchase Intention (Y) it is 1.3638. Since 1.3617 < 1.3638,
the data for the response variable (Purchase Intention) is slightly more dispersed than the explanatory
variable. Given the discrete nature of the 5-point Likert scale, there are no extreme mathematical outliers
in the raw data variables themselves.

3. SECTION 4.1: SCATTER DIAGRAM & CORRELATION

3.1 Scatter Plot

3.2 Pattern

The scatter plot reveals a positive, linear relationship between KOL Reputation and Purchase Intention.
As the rating for KOL reputation increases, there is a visible tendency for the purchase intention rating to
increase as well. The relationship appears to be of moderate strength.

3.3 Correlation

Correlation Coefficient (r): r = 0.5778

3.4 Test of Linear Relationship

To statistically verify the linear relationship, a hypothesis test is conducted:

• Null Hypothesis (H0): There is no linear relationship (ρ = 0).

The calculated P-value is 4.2997e-04. Since the P-value is significantly less than the standard
significance level (α = 0.05), we reject the null hypothesis. Conclusion: There is sufficient statistical
evidence to conclude that a significant positive linear relationship exists between KOL Reputation and
Purchase Intention.

4. SECTION 4.2: LEAST SQUARES REGRESSION

4.1 Regression Equation

The least squares regression line is formulated as:

ŷ = 0.9499 + 0.5787x

4.2 Interpretation

• Intercept (b0 = 0.9499): If a seller or KOL has a reputation score of 0, the predicted purchase
intention score is 0.9499. Since a score of 0 is outside the range of our 1-5 survey scale, the
intercept serves primarily as a mathematical anchor for the regression line rather than a practical
expected value.
• Slope (b1 = 0.5787): For every 1-unit increase in the KOL Reputation score, the estimated
Purchase Intention score increases, on average, by 0.5787 units.

4.3 Coefficient of Determination

tính R^2
R2: 0.3338 (or 33.38%)

Interpretation: Approximately 33.38% of the variation in the Purchase Intention (Y) can be explained by
the linear relationship with KOL Reputation (X). The remaining 66.62% of the variation is due to other
influential factors not included in this simple model (e.g., product price, brand loyalty, personal need).

4.4 Prediction

If a KOL receives a reputation rating of X = 4, the predicted Purchase Intention score is:
 ŷ = 0.9499 + 0.5787(4) = 3.2646

5. SECTION 4.3: RESIDUAL ANALYSIS

5.1 Residual Plot

The residual plot shows the residuals scattered somewhat randomly around the zero line. However,
because the data originates from discrete integer scales (1 to 5 Likert), the points form distinct linear
bands. Despite this structural artifact, there is no distinct curvature (like a U-shape), which supports the
use of a linear model.
5.2 Normality

The histogram of the residuals shows an approximately symmetric, bell-shaped distribution centered
near zero. This indicates that the normality assumption for the error terms is reasonably met, making
inference tests reliable.

5.3 Outliers

Analyzing the residuals, there are no points that exhibit severe deviation (e.g., standardized residuals
substantially beyond ±3). A few residuals approach ±2, which are typical expected variations in a sample
of this size. Therefore, no highly influential outliers need to be removed from this dataset.

5.4 Model Assumptions

• Linearity: The scatter plot and the lack of distinct curves in the residual plot validate the linearity
assumption.
• Independence: Assuming the survey respondents answered independently of one another, this
assumption holds.
• Constant Variance (Homoscedasticity): The residual plot shows a fairly consistent vertical
spread across the range of fitted values, satisfying the homoscedasticity assumption.
• Normality of Residuals: The histogram confirms an approximately normal distribution of errors.
6. DISCUSSION

The results align well with initial expectations: consumers are significantly more likely to purchase
products endorsed by KOLs or sellers with higher reputations. The positive correlation (r = 0.5778)
provides quantitative backing for influencer marketing strategies.

However, the model's explanatory power (R2 = 0.3338) suggests that KOL reputation is only one piece of
the puzzle. Other variables, such as product pricing, individual consumer needs, brand quality, and
promotional discounts, heavily influence the remaining 66.62% of the variance in purchase decisions. A
limitation of this study is the relatively small sample size (n = 33) and the restriction of data to integer-
based Likert scales, which causes data point overlapping in the regression analysis. Future research
should implement multiple regression models incorporating price sensitivity and demographic data to
create a more robust predictive model.

7. CONCLUSION

In summary, this research investigated the effect of KOL and seller reputation on customer purchase
intentions. The statistical analysis of 33 observations yielded a moderate positive linear relationship (r =
0.5778) and a valid linear regression model (ŷ = 0.9499 + 0.5787x). The data satisfies standard regression
assumptions, proving that reputation is a statistically significant predictor of buying behavior. Ultimately,
while reputation is not the sole determinant of a sale, it is undeniably a highly useful metric for
forecasting marketing success.
tóm tắt thêm
8. APPENDIX

Raw Data Summary

The dataset used for this analysis comprises 33 survey responses. (Full data available in the associated
Excel file).
ngọc minh làm lại