Pearson Correlation: Design Features & Performance

Welcome to the second module of our course, "Identifying Trends and Style Relationships." In the previous module, we focused on the essential groundwork of data preparation, culminating in a cleaned dataset of lingerie appeal scores. Now, with a trustworthy foundation in place, we can begin the exciting work of uncovering the patterns and relationships within that data.

This lesson introduces a fundamental tool for this discovery phase: the Pearson correlation coefficient. Our objective is to compute and interpret the correlation between a continuous design feature, like the depth of a bra's neckline, and a continuous performance metric, such as its photogenic appeal score. This will allow you to move from simple summary statistics to quantifying the linear association between specific design choices and their perceived success, a key step in your trend analysis.

From Clean Data to First Insights

You've already done the hard work of cleaning your data. The next logical question is, "What drives the appeal score?" Does more lace help? Does a wider strap detract? Pearson correlation is one of the first and most powerful tools to answer these kinds of questions, specifically when you are comparing two continuous variables.

It measures the strength and direction of a linear relationship. Since your background is in statistics, you're well-acquainted with the theory, so we'll focus directly on its application and interpretation in your design context.

For a quick refresher on the core concepts, the following resource provides a concise overview.

Pearson Correlation Coefficient: Calculation + Examples

This article from QuestionPro provides a clear, practical summary of Pearson correlation.

Please read the first two sections, "What is the Pearson correlation coefficient?" and "When to Use the Pearson Correlation Coefficient?". Pay attention to the distinction between positive and negative linear relationships.

In the world of apparel design, this isn't just a theoretical exercise. Researchers use correlation to validate new design processes. For instance, a study might investigate if the fit of a virtual 3D model correlates with the fit of the final physical garment. A strong positive correlation would mean the virtual prototyping is a reliable predictor of real-world fit, saving time and resources.

The following research paper does exactly that, using Pearson correlation to test their hypothesis.

[PDF] Relationship Between Virtual Prototype and Physical Fitting of an ...

This paper from the International Journal of Home Economics shows a direct application of Pearson correlation in an apparel context.

Focus on the "Result and Discussion" section. First, look at Table 3 and the accompanying text, which states the r-score. Then, review Table 4 and its explanation, where the authors compare their calculated r-value to a critical value to determine statistical significance and reject the null hypothesis. This is a classic application of the technique.

Numerical Exercise: Neckline Depth vs. Appeal Score

Let's apply this to your "Aphrodite" photoshoot project. Imagine you have data for 8 different plunge-style bras. You've measured the neckline depth (the vertical drop from the strap attachment point to the center gore) in centimeters, and you have the final appeal score (from 1-10) for each piece from your cleaned dataset.

Here is the data:

Piece ID	Neckline Depth (cm), $x$	Appeal Score (1-10), $y$
1	10	6
2	11	5
3	12	7
4	13	8
5	14	7
6	15	9
7	16	10
8	17	9

Your task is to calculate the Pearson correlation coefficient, $r$ , to determine if there is a linear relationship between neckline depth and appeal score.

To do this, we will use the standard formula for Pearson's $r$ :

r = \frac{N \sum x y - ( \sum x ) ( \sum y )}{[ N \sum x ^{2} - ( \sum x ) ^{2} ] [ N \sum y ^{2} - ( \sum y ) ^{2} ]}

Where $N$ is the number of data pairs (in this case, 8).

The following resource provides a step-by-step guide for this calculation, which we will follow.

Pearson Correlation Coefficient: Calculation + Examples

This section of the QuestionPro article lays out the formula and a clear, four-step process for the calculation.

Read the section "Pearson Correlation Coefficient Formula and Calculation". You will follow these exact steps for our numerical exercise.

Your Task:

Create a table with columns for $x$ , $y$ , $x y$ , $x^{2}$ , and $y^{2}$ .
Fill out the table for the 8 data points.
Calculate the sum of each column: $\sum x$ , $\sum y$ , $\sum x y$ , $\sum x^{2}$ , and $\sum y^{2}$ .
Substitute these sums into the Pearson correlation formula to find $r$ .

Click to see the detailed solution.

Step 1 & 2: Complete the data table.

Piece ID	$x$	$y$	$x y$	$x^{2}$	$y^{2}$
1	10	6	60	100	36
2	11	5	55	121	25
3	12	7	84	144	49
4	13	8	104	169	64
5	14	7	98	196	49
6	15	9	135	225	81
7	16	10	160	256	100
8	17	9	153	289	81
Sum	108	61	849	1500	485

Step 3: State the sums.

$N = 8$
$\sum x = 108$
$\sum y = 61$
$\sum x y = 849$
$\sum x^{2} = 1500$
$\sum y^{2} = 485$

Step 4: Plug the values into the formula.

r = \frac{8 ( 849 ) - ( 108 ) ( 61 )}{[ 8 ( 1500 ) - ( 108 ) ^{2} ] [ 8 ( 485 ) - ( 61 ) ^{2} ]}

r = \frac{6792 - 6588}{[ 12000 - 11664 ] [ 3880 - 3721 ]}

r = \frac{204}{[ 336 ] [ 159 ]}

r = \frac{204}{53424}

r = \frac{204}{231.136} \approx 0.8826

The Pearson correlation coefficient is approximately $r = 0.88$ .

Interpreting the Result

A numerical result is only useful when we can translate it back into a meaningful insight for our project. Our calculated $r \approx 0.88$ can be interpreted based on its direction and strength.

Pearson Correlation Coefficient: Calculation + Examples

These sections provide guidelines and visual aids for interpreting the correlation coefficient.

Start by reading about determining strength and how to interpret the coefficient. Then, review the definitions of Strength and Direction. Finally, look at the visual examples of scatterplots and the "rule of thumb" provided in the FAQs.

Based on these guidelines, here's how we interpret our finding:

Direction: The value is positive ( $0.88$ ). This indicates a positive correlation. In the context of our data, this means that as the neckline depth increases, the appeal score also tends to increase. Deeper plunge necklines are associated with higher appeal scores in this dataset.
Strength: A value of 0.88 is very close to 1. According to the common "rule of thumb" (e.g., 0.70-0.89 is 'strong', 0.90+ is 'very strong'), this indicates a strong positive linear relationship. The association between neckline depth and appeal score is not just random; it's quite consistent.

Pearson Correlation Coefficient: Calculation + Examples

A chart illustrating the interpretation of Pearson correlation coefficient values. Our result of 0.88 falls into the 'Strong positive relationship' category.

A Critical Caveat: Correlation ≠ Causation
This is a point your causal analysis background will appreciate. We found a strong positive correlation, but this does not mean that increasing neckline depth causes higher appeal scores. It's possible that a third, unobserved variable—a confounder—is responsible. For example, perhaps deeper necklines are only used with more luxurious fabrics (like silk charmeuse vs. standard satin), and it's the fabric, not the neckline, that is truly driving the appeal. We will explore how to untangle these effects in Module 5 when we delve into causal inference. For now, we note this correlation as a promising lead for our trend analysis.

Conclusion

In this lesson, you've taken the first step in analyzing the relationships within your data. You've learned how to apply the Pearson correlation coefficient to quantify the linear association between a continuous design feature and a performance metric.

Key Takeaways:

Pearson's $r$ is a value between -1 and +1 that measures the strength and direction of a linear relationship between two continuous variables.
A positive $r$ means as one variable increases, the other tends to increase. A negative $r$ means as one increases, the other tends to decrease.
Values closer to -1 or +1 indicate a stronger relationship, while values closer to 0 indicate a weaker one.
Always visualize your data with a scatterplot to confirm the linear assumption, and remember that correlation does not imply causation.

While Pearson correlation is excellent for linear relationships, many relationships in design and aesthetics are not perfectly linear. What if a style's appeal increases with a feature up to a point, and then plateaus? For this, we need a different tool. In our next lesson, we will explore Spearman rank correlation, which is designed to assess monotonic relationships, whether they are linear or not.

Recommended resources

From Clean Data to First Insights

Numerical Exercise: Neckline Depth vs. Appeal Score

Interpreting the Result

Conclusion