What does a PMCC value of $r = -0.95$ indicate about a data set?

It indicates a very strong negative linear correlation. As one variable increases, the other variable decreases in a manner that very closely follows a straight line.

How do you distinguish between a Power model and an Exponential model using graphs?

A Power model ($y=ax^n$) becomes linear when plotting $\log y$ against $\log x$. An Exponential model ($y=kb^x$) becomes linear when plotting $\log y$ against $x$.

Why is it dangerous to use a regression model to predict values far outside the observed data range?

This is called extrapolation. It is unreliable because there is no evidence that the observed relationship (linear or non-linear) continues to hold true beyond the measured intervals.

If the linearized regression line is $Y = 0.5X + 2$ where $Y = \log_{10} y$ and $X = \log_{10} x$, what is the original model?

The model is a Power model $y = ax^n$. Here, $n = 0.5$ and $\log_{10} a = 2$, so $a = 10^2 = 100$. The equation is $y = 100x^{0.5}$.

What is the difference between $r=1$ and a high positive gradient in a linear regression?

$r=1$ means the points fall exactly on a line, indicating perfect correlation. The gradient tells you the rate of change between variables, but does not affect the PMCC value itself.

What error is made if a student calculates a PMCC of $1.2$?

The PMCC must always be between $-1$ and $1$. A value of $1.2$ indicates a calculation error, likely in the covariance or standard deviation components of the formula.

In the transformation of $y = kb^x$, what does the gradient of the resulting line represent?

The gradient of the line (when plotting $\log y$ vs $x$) represents $\log b$. To find the original constant $b$, you must calculate the inverse log of the gradient.

Does a PMCC of $0$ mean there is no relationship between the variables?

No, it only means there is no **linear** relationship. The variables could still have a perfect non-linear relationship, such as a quadratic or exponential curve.

What is 'coding' in the context of non-linear regression?

Coding is the process of transforming the original data (e.g., by taking logarithms) to change a non-linear relationship into a linear one so that standard regression techniques can be applied.

When decoding the intercept $c$ to find constant $a$ in $y=ax^n$, what must you check first?

You must check the base of the logarithm used for the transformation. If $\ln$ was used, $a = e^c$; if $\log_{10}$ was used, $a = 10^c$.

Library Podcasts

Courses

Referral & Rewards

Data Presentation & Interpretation

PMCC & Non-linear Regression

Summary

This topic explores the numerical measurement of linear relationships through the Product Moment Correlation Coefficient (PMCC) and the methodology for analyzing non-linear bivariate data using logarithmic transformations to create linear models.

1. Product Moment Correlation Coefficient (PMCC)

Definition: The PMCC, denoted by $r$ , is a statistical measure that quantifies the strength and direction of a linear correlation between two variables in a bivariate data set.
Numerical Range: The value of $r$ is strictly bounded such that $-1 \le r \le 1$ . A value of $1$ indicates perfect positive linear correlation, while $-1$ indicates perfect negative linear correlation.
Interpretation of Strength: The closer the absolute value $|r|$ is to $1$ , the stronger the linear relationship. An $r$ value of $0$ suggests that there is no linear correlation between the variables, though a non-linear relationship might still exist.
Invariance: The PMCC is independent of the gradient of the line of best fit; it only measures how closely the data points cluster around a straight line.

A scatter plot showing a curved, non-linear relationship between two variables, illustrating why linear regression alone is insufficient.

2. Non-linear Regression Models

3. Data Coding via Logarithms

4. Key Distinctions: Model Selection

5. Methods & Techniques: Solving Problems

6. Exam Strategy & Common Pitfalls