Welcome to Interpreting Lines of Best Fit and Correlation Coefficients

Have you ever noticed that as the temperature rises in the summer, ice cream sales go up too? Or that the more hours you spend revising, the higher your test scores tend to be? In statistics, when we examine two variables at the same time, we call it bivariate data.

In this chapter, you will learn how to make sense of relationships between two variables. We will explore how to draw and interpret a line of best fit, how to make sensible predictions, how to measure the strength of a relationship using correlation coefficients, and why you must never confuse correlation with cause and effect. Don't worry if this sounds like a lot right now — we will break everything down step-by-step!

1. Scatter Diagrams and the Line of Best Fit

A scatter diagram (or scatter graph) is used to display bivariate data. Each point represents a pair of numbers: one on the horizontal axis and one on the vertical axis.

Understanding the Two Variables

1. Explanatory (Independent) Variable: Plotted on the horizontal \(x\)-axis. This is the variable that you control or think is causing a change in the other.

2. Response (Dependent) Variable: Plotted on the vertical \(y\)-axis. This is the outcome variable that you measure.

Example: If you want to see how the amount of study time affects an exam score, Study Time is the explanatory variable (\(x\)) and Exam Score is the response variable (\(y\)).

The Mean Point Rule: How to Draw the Line of Best Fit

A line of best fit (also called a linear trend line or regression line) is a single straight line that shows the general direction of the data points. In your CCEA GCSE Statistics exam, there is a golden rule for drawing this line accurately:

Step 1: Calculate the mean of all the \(x\)-values, written as \(\bar{x}\) (read as "x-bar").
Step 2: Calculate the mean of all the \(y\)-values, written as \(\bar{y}\) (read as "y-bar").
Step 3: Plot the mean point \((\bar{x}, \bar{y})\) on your scatter graph. Mark it clearly with a small cross or label.
Step 4: Using a ruler, draw a single straight line that passes directly through the mean point \((\bar{x}, \bar{y})\), balancing the remaining data points so that roughly equal numbers of points lie above and below the line.

Memory Tip: Never join the dots to make a zig-zag! A line of best fit is always one continuous straight line that must pass through the double mean point \((\bar{x}, \bar{y})\).

The Equation of a Line of Best Fit: \(y = mx + c\)

Because the line of best fit is a straight line, it can be written algebraically in the form:

\(y = mx + c\)    (or \(y = ax + b\))

In this equation:
The Gradient (\(m\) or \(a\)): Represents the estimated rate of change. It tells you by how much the response variable (\(y\)) changes for every \(1\) unit increase in the explanatory variable (\(x\)).
The \(y\)-intercept (\(c\) or \(b\)): Represents the estimated baseline value of \(y\) when \(x = 0\).

Worked Example:
Suppose the line of best fit for a plant's height (\(y\) in cm) after \(x\) weeks is given by the equation \(y = 3.5x + 4\).
Interpretation of Gradient (\(3.5\)): On average, the plant grows by \(3.5\text{ cm}\) each week.
Interpretation of Intercept (\(4\)): The initial height of the plant at week \(0\) was \(4\text{ cm}\).

Key Takeaway for Section 1: Always calculate and plot the mean point \((\bar{x}, \bar{y})\) before drawing your line of best fit. The gradient tells you the rate of change per unit of \(x\), and the intercept gives the starting value when \(x = 0\).

2. Making Predictions: Interpolation vs Extrapolation

Once you have a line of best fit (or its equation), you can use it to estimate values. However, not all estimates are equally reliable!

Interpolation

Interpolation means estimating an unknown value inside the range of your collected sample data.

Reliability: Estimates made by interpolation are generally reliable, provided there is a strong linear correlation and enough data points.

Example: If you collected data on students who studied between \(1\) hour and \(8\) hours, estimating the exam score for a student who studied \(4\) hours is interpolation (since \(4\) lies between \(1\) and \(8\)).

Extrapolation

Extrapolation means predicting a value outside the range of your observed data by extending the line of best fit beyond your actual data points.

Reliability: Estimates made by extrapolation are unreliable. You cannot assume that the same linear trend continues beyond the range of your collected data.

Example: Using the same data (\(1\) to \(8\) hours of study), estimating the score for someone who studied \(25\) hours is extrapolation. A person cannot study \(25\) hours in a single day, and exam scores cannot exceed \(100\%\)!

A Golden Rule for GCSE Statistics Predictions

When using a regression line of \(y\) on \(x\), you should only use a given value of \(x\) to predict a value of \(y\). You must not work backwards to predict an \(x\)-value from a given \(y\)-value using a line of \(y\) on \(x\).

Key Takeaway for Section 2: Interpolation is inside the data range (reliable). Extrapolation is outside the data range (unreliable because the trend may change).

3. Correlation and Correlation Coefficients

Correlation describes the nature and strength of the linear relationship between two variables.

Types of Correlation

Positive Correlation: As \(x\) increases, \(y\) increases (the line slopes upwards from left to right). Example: Height and shoe size.
Negative Correlation: As \(x\) increases, \(y\) decreases (the line slopes downwards from left to right). Example: Outside temperature and heating bills.
Zero / No Linear Correlation: There is no apparent linear pattern between the two variables. Example: Shoe size and test score in mathematics.
Non-Linear (Curved) Relationship: The points follow a clear curve (like a parabola or wave). Even though there is a clear pattern, a standard linear correlation coefficient will be close to \(0\) because the relationship is not a straight line.

The Correlation Coefficient Scale

We measure correlation using numerical values called correlation coefficients. Two common types are Pearson’s product-moment correlation coefficient (\(r\)) and Spearman’s rank correlation coefficient (\(r_s\)).

Both correlation coefficients always fall within the range:

\(-1 \le r \le 1\)

Here is how to interpret the numerical values:
\(r = +1\): Perfect positive linear correlation (all points lie exactly on an upward straight line).
\(r = -1\): Perfect negative linear correlation (all points lie exactly on a downward straight line).
\(r = 0\): No linear correlation.

Guidelines for Describing Strength

In your exam, you should describe both the strength and the direction of the correlation using these qualitative guidelines:

Strong Correlation: \(|r| \ge 0.8\)   (e.g. \(r = +0.85\) is strong positive; \(r = -0.92\) is strong negative)
Moderate Correlation: \(0.5 \le |r| < 0.8\)   (e.g. \(r = +0.65\) is moderate positive; \(r = -0.58\) is moderate negative)
Weak Correlation: \(0 < |r| < 0.5\)   (e.g. \(r = +0.25\) is weak positive; \(r = -0.31\) is weak negative)

Higher Tier: Spearman’s Rank Correlation Coefficient (\(r_s\))

For Higher Tier candidates, you may need to calculate and interpret Spearman’s rank correlation coefficient (\(r_s\)). This measures how well the relationship between two variables can be described using a monotonic (ordered) function.

The formula (given on your examination formula sheet) is:

\(r_s = 1 - \frac{6 \sum d^2}{n(n^2 - 1)}\)

Where:
\(d\) = the difference between the ranks for each pair of data items.
\(\sum d^2\) = the sum of the squared differences.
\(n\) = the number of pairs of data items.

Key Takeaway for Section 3: Correlation coefficients range from \(-1\) to \(+1\). The sign (\(+\) or \(-\)) gives the direction, while the number's distance from \(0\) gives the strength.

4. Correlation vs Causation

One of the most important lessons in statistics is:

Correlation does NOT imply Causation!

Just because two variables show a strong statistical link does not mean that changes in one variable actually cause changes in the other.

Confounding (Lurking) Variables

Often, an apparent relationship between two variables is caused by a third, unmeasured factor called a confounding variable (or lurking variable).

The Classic Ice Cream Analogy:
In summer towns, statistics show a strong positive correlation between daily ice cream sales and the number of drownings.
Does eating ice cream cause people to drown? Of course not!
The Confounding Variable: Hot Weather. When the temperature is hot, more people buy ice cream AND more people go swimming in the sea, leading to more drownings. The temperature is the lurking variable driving both.

The Impact of Outliers (Anomalies)

An outlier is a data point that lies far away from the overall pattern of the rest of the data. Outliers can:

1. Pull the line of best fit towards them (exerting "leverage"), distorting the true trend.
2. Artificially weaken or inflate the value of the correlation coefficient (\(r\)).

Always inspect scatter diagrams for outliers before drawing conclusions!

Key Takeaway for Section 4: When two variables correlate, never write that one "causes" the other in exam answers. Use words like "associated with" or "related to", and consider possible confounding variables.

5. Common Pitfalls & Examiner Tips

Avoid these common mistakes in your CCEA GCSE Statistics exam:

Mistake 1: Confusing Gradient with Strength of Correlation
Wrong: "This graph has a strong correlation because the line is very steep."
Right: Correlation strength depends entirely on how closely the data points cluster around the line, not on how steep the line is.

Mistake 2: Missing the Mean Point
Wrong: Drawing a line of best fit by connecting the first and last points on the graph.
Right: Calculate \((\bar{x}, \bar{y})\), plot it, and make sure your ruler passes through that exact coordinate.

Mistake 3: Stating That \(r = 0\) Means "No Relationship"
Wrong: "\(r = 0\) means there is no relationship at all between the variables."
Right: \(r = 0\) means there is no linear (straight-line) relationship. A strong curved (non-linear) relationship could still exist!

Mistake 4: Claiming Cause and Effect
Wrong: "Increasing revision causes exam marks to go up."
Right: "There is a strong positive correlation between revision time and exam marks."

Mistake 5: Unjustified Extrapolation
Wrong: Using the line to predict a value far outside the data and stating it is completely accurate.
Right: Stating the prediction, but adding that it is extrapolation and therefore unreliable because the trend may not continue.

6. Quick Revision Checklist

Before sitting your exam, check that you can:

✓ Calculate the double mean point \((\bar{x}, \bar{y})\) and use it to draw an accurate line of best fit.
✓ Interpret the gradient \(m\) and intercept \(c\) in the linear model \(y = mx + c\) using the real-world context of the question.
✓ Distinguish between interpolation (reliable, inside range) and extrapolation (unreliable, outside range).
✓ Describe correlation by both its direction (positive/negative) and strength (strong/moderate/weak).
✓ Know that correlation coefficients lie between \(-1\) and \(+1\).
✓ Explain why correlation does not prove causation by identifying potential confounding variables.
✓ (Higher Tier) Apply the formula for Spearman’s rank correlation coefficient: \(r_s = 1 - \frac{6 \sum d^2}{n(n^2 - 1)}\).