Welcome to Bivariate Analysis!
In statistics, single-variable data (univariate data) tells us about one thing at a time, like the heights of students in a class. But what if we want to explore whether taller students also have longer arm spans? Or whether spending more hours revising leads to higher test scores?
That is where bivariate analysis comes in! The prefix bi- means two (think of a bicycle with two wheels). Bivariate data is simply numerical data that involves two variables measured for each individual or item.
In this chapter for CCEA GCSE Further Mathematics (Unit 3: Statistics), you will learn how to display two variables visually, draw and calculate the line of best fit, evaluate predictions, and calculate Spearman’s Rank Correlation Coefficient. Let's break it all down step by step!
1. Scatter Graphs & Correlation
A scatter graph (or scatter diagram) provides a visual representation of bivariate data. Each point plotted represents a single subject with two recorded values.
Independent vs. Dependent Variables
When drawing a scatter graph, the axes have specific roles:
• The \(x\)-axis (Horizontal): Holds the explanatory (independent) variable. This is the variable you control or suspect is causing an effect.
• The \(y\)-axis (Vertical): Holds the response (dependent) variable. This is the outcome or measurement you are observing.
Types of Linear Correlation
Correlation describes the statistical relationship or association between two numerical variables:
• Positive Correlation: As the values of \(x\) increase, the values of \(y\) tend to increase (the points drift upwards from left to right). Example: Temperature and ice cream sales.
• Negative Correlation: As the values of \(x\) increase, the values of \(y\) tend to decrease (the points drift downwards from left to right). Example: Car age and resale value.
• Zero / No Correlation: There is no apparent linear relationship between \(x\) and \(y\). The points appear randomly scattered. Example: Shoe size and GCSE Maths score.
Strength of Correlation
We also describe how tightly the data points cluster around a straight line:
• Strong Correlation: The plotted points lie very close to a clear straight line.
• Moderate Correlation: The plotted points show an identifiable linear trend, but with more spread.
• Weak Correlation: The plotted points show only a slight linear trend with a wide spread.
Crucial Exam Concept: Correlation vs. Causation
Always remember: Correlation does NOT imply causation!
Just because two variables show a strong positive or negative correlation does not mean that changes in one variable directly cause changes in the other. There may be a hidden third factor, or the link may be purely coincidental.
Key Takeaway for Section 1: Scatter graphs plot \(x\) (explanatory) against \(y\) (response). Correlation can be positive, negative, or zero, and its strength can be strong, moderate, or weak. Correlation shows an association, not a cause!
2. The Mean Point and the Line of Best Fit
A line of best fit is a straight line drawn through the plotted points to model the general trend of the data.
Step 1: Calculating the Mean Point \((\bar{x}, \bar{y})\)
In CCEA GCSE Further Mathematics, your line of best fit must always pass directly through the mean point, denoted as \((\bar{x}, \bar{y})\). You calculate the coordinates using the formulae:
\(\bar{x} = \frac{\sum x}{n}\) and \(\bar{y} = \frac{\sum y}{n}\)
where \(n\) is the total number of data pairs, and \(\sum\) represents the sum of the values.
Step 2: Drawing the Line of Best Fit Accurately
Follow these rules when drawing the line on an exam paper:
1. Plot the mean point \((\bar{x}, \bar{y})\) accurately on your scatter diagram.
2. Place your ruler so it goes exactly through the mean point.
3. Rotate the ruler so it follows the overall angle/trend of the points, ensuring an even balance of points above and below the line.
4. Draw a single, continuous, straight line using a sharp pencil.
Step 3: Finding the Equation of the Line of Best Fit
You may be asked to determine the linear equation in the form:
\(y = mx + c\) or \(y = a + bx\)
Here is the foolproof step-by-step method:
1. Find the Gradient (\(m\) or \(b\)):
Pick two points on the drawn line itself: \((x_1, y_1)\) and \((x_2, y_2)\). Do not simply pick original data points from the table unless they happen to lie directly on your line!
\(m = \frac{y_2 - y_1}{x_2 - x_1}\)
2. Find the \(y\)-intercept (\(c\) or \(a\)):
Substitute your calculated gradient \(m\) and the mean point \((\bar{x}, \bar{y})\) into the straight-line equation:
\(c = \bar{y} - m\bar{x}\)
3. Write out the final equation: State \(y = mx + c\) using your calculated numerical values.
Key Takeaway for Section 2: Calculate \((\bar{x}, \bar{y})\) first. Your line of best fit must pass through this coordinate. Always select coordinates from the drawn line to calculate the gradient.
3. Making Predictions: Interpolation vs. Extrapolation
Once you have a line of best fit (or its equation), you can use it to estimate the value of one variable given the value of the other.
Interpolation
Interpolation means estimating a value within the range of your collected data points (between the smallest and largest observed \(x\)-values).
• Reliability: Generally reliable, provided the scatter graph shows a strong or moderate linear correlation.
Extrapolation
Extrapolation means estimating a value outside the range of your collected data points (smaller than the minimum observed value or larger than the maximum observed value).
• Reliability: Inherently unreliable. We have no evidence that the same linear trend continues beyond our observed data boundaries.
Exam Tip: If a question asks, "Comment on the reliability of this estimate," check whether the value is inside the given data range (interpolated) or outside the data range (extrapolated), and state whether the correlation is strong or weak!
Key Takeaway for Section 3: Estimating inside the data range = Interpolation (reliable). Estimating outside the data range = Extrapolation (unreliable).
4. Spearman’s Rank Correlation Coefficient (\(r_s\))
While a scatter graph provides a visual check of correlation, Spearman’s Rank Correlation Coefficient (\(r_s\)) provides a precise mathematical measure of the agreement between the rankings of two variables.
The Formula
This formula is provided on your CCEA GCSE Further Mathematics Formula Sheet:
\(r_s = 1 - \frac{6 \sum d^2}{n(n^2 - 1)}\)
where:
• \(n\) = number of pairs of data
• \(d\) = difference between the ranks for each pair: \(d = \text{Rank}(x) - \text{Rank}(y)\)
• \(\sum d^2\) = the sum of the squared rank differences
Step-by-Step Method to Calculate \(r_s\)
Step 1: Rank the first variable (\(x\))
Assign rank 1 to either the highest or lowest value, rank 2 to the next, and so on. Be consistent!
Step 2: Rank the second variable (\(y\))
Rank the \(y\) values using the exact same direction (e.g., if rank 1 was the highest score for \(x\), rank 1 must also be the highest score for \(y\)).
Step 3: Handle any tied ranks (Crucial!)
When two or more values are identical, assign each tied value the mean (average) of the rank positions they would have occupied.
Example: If two items tie for 3rd and 4th place, they share ranks 3 and 4. You give both items rank \(\frac{3 + 4}{2} = 3.5\). The next item receives rank 5.
Step 4: Calculate the difference \(d\)
For each pair of ranks, calculate \(d = \text{Rank}(x) - \text{Rank}(y)\).
Sanity check: The sum of differences \(\sum d\) should always equal 0 (or very close to 0 if there are rounded ties).
Step 5: Square each difference (\(d^2\)) and sum them
Calculate \(d^2\) for each row and find \(\sum d^2\). Note that squared numbers are always non-negative!
Step 6: Substitute into the formula
Carefully calculate \(n(n^2 - 1)\), multiply \(\sum d^2\) by 6, divide, and subtract from 1.
Interpreting \(r_s\) Values
The value of \(r_s\) always lies between \(-1\) and \(+1\) inclusive:
\(-1 \le r_s \le 1\)
• \(r_s = +1\): Perfect positive rank correlation (rankings agree completely).
• \(r_s = 0\): No rank correlation (no association between rankings).
• \(r_s = -1\): Perfect negative rank correlation (rankings are in exact reverse order).
Key Takeaway for Section 4: Rank both variables consistently, handle ties by averaging the positions, square each difference \(d\), and compute \(r_s = 1 - \frac{6 \sum d^2}{n(n^2 - 1)}\).
5. Worked Example: Calculating Spearman's Rank
Five students (\(n = 5\)) completed a Physics test (\(x\)) and a Mathematics test (\(y\)):
• Student A: Physics = 42, Maths = 50
• Student B: Physics = 65, Maths = 60
• Student C: Physics = 78, Maths = 75
• Student D: Physics = 65, Maths = 70
• Student E: Physics = 90, Maths = 85
Step 1 & 2: Ranking the Data (Highest value = Rank 1)
Physics scores: 90 (Rank 1), 78 (Rank 2), 65 & 65 (Tied for 3rd & 4th \(\implies \frac{3+4}{2} = 3.5\)), 42 (Rank 5).
Maths scores: 85 (Rank 1), 75 (Rank 2), 70 (Rank 3), 60 (Rank 4), 50 (Rank 5).
Step 3 & 4: Differences and Squared Differences
• Student A: Rank(\(x\)) = 5, Rank(\(y\)) = 5 \(\implies d = 0 \implies d^2 = 0\)
• Student B: Rank(\(x\)) = 3.5, Rank(\(y\)) = 4 \(\implies d = -0.5 \implies d^2 = 0.25\)
• Student C: Rank(\(x\)) = 2, Rank(\(y\)) = 2 \(\implies d = 0 \implies d^2 = 0\)
• Student D: Rank(\(x\)) = 3.5, Rank(\(y\)) = 3 \(\implies d = 0.5 \implies d^2 = 0.25\)
• Student E: Rank(\(x\)) = 1, Rank(\(y\)) = 1 \(\implies d = 0 \implies d^2 = 0\)
Sum of squared differences: \(\sum d^2 = 0 + 0.25 + 0 + 0.25 + 0 = 0.5\)
Step 5: Apply the Formula
\(n = 5 \implies n(n^2 - 1) = 5(5^2 - 1) = 5(24) = 120\)
\(r_s = 1 - \frac{6 \times 0.5}{120} = 1 - \frac{3}{120} = 1 - 0.025 = 0.975\)
Interpretation: An \(r_s\) value of \(0.975\) indicates a very strong positive correlation between rankings in Physics and Mathematics.
6. Common Exam Traps & How to Avoid Them
CCEA examiners frequently report the same preventable errors year after year. Watch out for these:
• Trap 1: Line misses the mean point. Always plot \((\bar{x}, \bar{y})\) clearly and ensure your ruler passes right through it before drawing your line of best fit.
• Trap 2: Inconsistent ranking directions. Never rank \(x\) from highest to lowest while ranking \(y\) from lowest to highest. Stick to the same rule for both variables.
• Trap 3: Tied rank slips. If two values tie for positions 4 and 5, assign both \(4.5\). The next item must be assigned rank 6, NOT rank 5!
• Trap 4: Calculator order of operations. In Spearman's formula, calculate \(\frac{6\sum d^2}{n(n^2-1)}\) first, and then subtract the result from \(1\).
• Trap 5: Gradient points from the raw table. When finding the equation \(y = mx + c\), read two coordinate pairs directly off the drawn line, not from the original data table.
Quick Summary Checklist
Before sitting your Unit 3 exam, make sure you can:
✓ State whether a scatter graph displays positive, negative, or zero correlation.
✓ Calculate the mean point \((\bar{x}, \bar{y})\) and draw the line of best fit through it.
✓ Find the gradient \(m\) from the line and calculate the \(y\)-intercept \(c = \bar{y} - m\bar{x}\).
✓ Distinguish between reliable interpolation and unreliable extrapolation.
✓ Rank data correctly (including tied ranks) and calculate \(r_s = 1 - \frac{6\sum d^2}{n(n^2-1)}\).
✓ Correctly interpret \(r_s\) values between \(-1\) and \(+1\).