PsychologyCorrelation, Hypothesis TestingUniversity

AQAAPOntarioNSWCBSEGCE O-LevelMoECAPS

Fisher's Z-transformation

Transforms Pearson's correlation coefficient (r) into a normally distributed variable (z').

Understand the formulaSee the free derivationOpen the full walkthrough

z^{'} = 0.5 ln (\frac{1 + r}{1 - r})

Open Full Walkthrough Try Calculator

This public page keeps the free explanation visible and leaves premium worked solving, advanced walkthroughs, and saved study tools inside the app.

Core idea

Overview

Fisher's Z-transformation converts Pearson's correlation coefficient (r) into a new variable, z', which is approximately normally distributed. This transformation is crucial for statistical inference involving correlation coefficients, especially when comparing correlations from different samples or constructing confidence intervals. It addresses the non-normal sampling distribution of r, particularly for extreme values, making it suitable for parametric tests.

When to use: Use this transformation when you need to perform hypothesis tests or construct confidence intervals for Pearson's correlation coefficient, especially when comparing correlations between two independent samples or when the sample size is small and r is far from zero. It's also used to average correlation coefficients.

Why it matters: This transformation is vital in psychological research for accurately comparing the strength of relationships between variables across different studies or groups. It allows researchers to apply standard statistical methods that assume normality, leading to more robust and reliable conclusions about correlations, which are fundamental to understanding psychological phenomena.

Symbols

Variables

r = Pearson's Correlation Coefficient, z' = Fisher's Z-score

r

Pearson's Correlation Coefficient

Variable

Fisher's Z-score

Variable

Walkthrough

Derivation

Formula: Fisher's Z-transformation

Fisher's Z-transformation converts Pearson's correlation coefficient into a normally distributed variable for statistical inference.

The input 'r' is a Pearson product-moment correlation coefficient.
The sample from which 'r' is derived is sufficiently large for the approximation to hold.
The variables being correlated are approximately bivariate normal.

Problem with 'r' Distribution:

The sampling distribution of Pearson's correlation coefficient 'r' is not normal, especially when the true population correlation is far from zero or with small sample sizes. This non-normality makes standard parametric tests (like t-tests or z-tests) inappropriate for 'r' directly.

r \sim skewed for extreme values

Note: The distribution of 'r' is symmetric only when the population correlation is zero.

Introducing the Transformation:

Ronald Fisher proposed this transformation to convert 'r' into a new variable, z', whose sampling distribution is approximately normal, regardless of the true population correlation or sample size (for sufficiently large N). The natural logarithm (ln) is used to 'stretch' the tails of the distribution.

z^{'} = 0.5 ln (\frac{1 + r}{1 - r})

Note: This transformation is also known as the arc-hyperbolic tangent (arctanh) function: z' = arctanh(r).

Properties of z':

The transformed variable z' has an approximately normal distribution with a mean of 0.5 * ln((1+ρ)/(1-ρ)) (where ρ is the population correlation) and a variance of approximately 1/(N-3), where N is the sample size. This stable variance simplifies hypothesis testing and confidence interval construction.

Var (z^{'}) \approx \frac{1}{N - 3}

Note: The standard error of z' is approximately 1/√(N-3).

Result

Var (z^{'}) \approx \frac{1}{N - 3}

Source: Fisher, R. A. (1921). On the 'probable error' of a coefficient of correlation deduced from a small sample. Metron, 1(4), 3-32.

Free formulas

Rearrangements

Solve for $r$

Fisher's Z-transformation: Make r the subject

r = \frac{e ^{2 z^{'}} - 1}{e ^{2 z^{'}} + 1}

To make $r$ the subject of Fisher's Z-transformation, one must use algebraic manipulation involving exponentiation to reverse the logarithmic transformation.

Difficulty: 4/5

The static page shows the finished rearrangements. The app keeps the full worked algebra walkthrough.

Visual intuition

Graph

The graph follows a curve that grows rapidly as the correlation coefficient r approaches one and decreases toward negative infinity as r approaches negative one, with vertical asymptotes at these boundaries because the function is undefined outside this range. For a psychology student, this shape means that the transformation stretches the scale of correlation coefficients, making differences between high values of r more pronounced than differences between values near zero. The most important feature is that the curve is steepest near the boundaries of negative one and one, which demonstrates that the transformation is most sensitive to changes in correlation when the relationship between variables is already very strong.

Graph type: logarithmic

Why it behaves this way

Intuition

Imagine a flexible ruler representing the correlation coefficient 'r', where the ends are squashed and the middle is stretched. Fisher's Z-transformation reshapes this ruler into a perfectly straight, infinitely long

The Fisher's Z-transformed value of Pearson's correlation coefficient (r).

This is a re-scaled version of 'r' that has an approximately normal sampling distribution, making it suitable for standard statistical tests.

r

Pearson's product-moment correlation coefficient, quantifying the linear relationship between two variables.

Indicates the strength and direction of a straight-line association; +1 for perfect positive, -1 for perfect negative, 0 for no linear relationship.

0.5

A constant scaling factor applied to the natural logarithm.

It's part of the mathematical definition to ensure the transformed variable 'z'' has the desired statistical properties, particularly its variance.

ln

The natural logarithm function.

This function transforms the bounded, non-linear scale of 'r' into an unbounded, more linear scale for 'z'', which is essential for approximating a normal distribution.

\frac{1 + r}{1 - r}

A ratio derived from 'r' that maps the range of 'r' (-1 to 1) to a range from 0 to infinity.

This ratio stretches the extreme values of 'r' and compresses values near zero, preparing the data for the logarithmic transformation to achieve approximate normality.

Signs and relationships

\ln(...): The natural logarithm is used to transform the bounded range of 'r' (-1 to 1) into an unbounded range for 'z'' (-∞ to +∞), which is a characteristic of a normal distribution.
\frac{1+r}{1-r}: This specific ratio is designed to stretch the tails of the 'r' distribution. As 'r' approaches 1, the denominator (1-r) approaches 0, making the ratio very large. As 'r' approaches -1, the numerator (1+r)

Free study cues

Insight

Canonical usage

This equation transforms a dimensionless Pearson's correlation coefficient into another dimensionless variable for statistical analysis.

Common confusion

Students sometimes mistakenly attempt to assign physical units to correlation coefficients or their transformations. However, these are statistical measures and are inherently dimensionless.

Dimension note

Both Pearson's correlation coefficient (r) and its Fisher Z-transformation (z') are dimensionless quantities. They are pure numbers that quantify the strength and direction of a statistical relationship between two

Unit systems

$r$ dimensionless · Pearson's correlation coefficient, a pure number representing the strength and direction of a linear relationship, always falls between -1 and +1.

z'dimensionless · Fisher's Z-transformed value, a pure number used for statistical inference, which is approximately normally distributed.

Ballpark figures

Quantity:

One free problem

Practice Problem

A researcher finds a Pearson correlation coefficient of r = 0.6 between two variables. Calculate the Fisher's Z-transformation (z') for this correlation.

Pearson's Correlation Coefficient0.6

Solve for: z'

Hint: Use the natural logarithm function.

The full worked solution stays in the interactive walkthrough.

Where it shows up

Real-World Context

Comparing the strength of correlation between anxiety and academic performance in two different student populations.

Study smarter

Tips

Ensure 'r' is a Pearson product-moment correlation coefficient.
The transformation makes the sampling distribution of z' approximately normal.
Remember to transform z' back to r when interpreting results (using the inverse transformation).
This transformation is particularly useful for small sample sizes or when 'r' is close to +1 or -1.

Avoid these traps

Common Mistakes

Applying the transformation to non-Pearson correlation coefficients.
Forgetting to convert z' back to r for interpretation.
Misinterpreting z' as a standard Z-score from a normal distribution.

Common questions

Frequently Asked Questions

Fisher's Z-transformation converts Pearson's correlation coefficient into a normally distributed variable for statistical inference.

Use this transformation when you need to perform hypothesis tests or construct confidence intervals for Pearson's correlation coefficient, especially when comparing correlations between two independent samples or when the sample size is small and r is far from zero. It's also used to average correlation coefficients.

This transformation is vital in psychological research for accurately comparing the strength of relationships between variables across different studies or groups. It allows researchers to apply standard statistical methods that assume normality, leading to more robust and reliable conclusions about correlations, which are fundamental to understanding psychological phenomena.

Applying the transformation to non-Pearson correlation coefficients. Forgetting to convert z' back to r for interpretation. Misinterpreting z' as a standard Z-score from a normal distribution.