Cohen's D Helped Coefficient Clarity

By Grace

Cohen's d, an effect size measure developed by the influential statistician Jacob Cohen, has gained prominence in research circles as a robust tool for interpreting and comparing the results of statistical tests. It's often used in settings where the F or t-statistic isn't sufficient for discerning practical significance, offering a way to quantify the strength of evidence by examining the standardized difference between two means. Let's delve into the origins, key aspects, and applications of Cohen's d, as laid out in Cohen's original paper.

a newspaper article with an image of a man singing into a microphone and holding a guitar
a newspaper article with an image of a man singing into a microphone and holding a guitar

Cohen, in his seminal 1962 paper "A Co Saint-Berthélemi le-roi II of scientific progress?", introduced the concept of effect sizes, arguing that statistical significance testing alone wasn't adequate for drawing conclusions about the practical significance of research findings. He proposed various effect size measures, among them Cohen's d, which is the focus of this discussion.

an old book page with words written in black and white on the bottom half of it
an old book page with words written in black and white on the bottom half of it

The Origin and Formula of Cohen's d

Cohen's d is a simple yet powerful statistic, defined as the difference between two means divided by the standard deviation of the data. In his original paper, Cohen represented the formula as:

an old book with writing on it and the title in english, as well as other words
an old book with writing on it and the title in english, as well as other words

d = (M1 - M2) / s

where M1 and M2 are the means of the two groups, and s is the standard deviation of the pooled data.

Papercraft KING FISHER
Papercraft KING FISHER

Interpreting Cohen's d

Cohen provided helpful guidelines for interpreting d values, which have since become widely accepted:

  • Small effect: d = 0.2
  • Medium effect: d = 0.5
  • Large effect: d = 0.8
an image of a man with books above his head and ink pen in the air
an image of a man with books above his head and ink pen in the air

Variations of Cohen's d

While the original formula is suitable for scenarios where sample sizes are large and similar, various extensions and modifications have been proposed for other scenarios:

  • Hedge's g: A modified version that adjusts for bias in small samples.
  • Mean-difference d: A variation that uses the standard deviation of the control group as the denominator.
an old piece of paper with writing on it
an old piece of paper with writing on it

Beyond the Basics: Assumptions and Limitations

While Cohen's d offers a great deal of utility, it's essential to understand its underlying assumptions and limitations:

an old book with cursive writing on the pages and numbers in each section
an old book with cursive writing on the pages and numbers in each section
white crumpled paper textured background with space for your text or image, high resolution photo
white crumpled paper textured background with space for your text or image, high resolution photo
an image of a white background with many words in the same font and numbers on it
an image of a white background with many words in the same font and numbers on it
k
k
cohen's d original paper
cohen's d original paper
🟢
🟢
white crumpled paper textured with some sort of pattern
white crumpled paper textured with some sort of pattern
a paper box that is open to show the front and back side of it, with an image of a cartoon character on top
a paper box that is open to show the front and back side of it, with an image of a cartoon character on top
a white sheet of paper with dots on it
a white sheet of paper with dots on it

Assumptions

  • The data is approximately normally distributed.
  • The group variances are equal.

Limitations

Cohen's d is sensitive to violations of its assumptions, particularly non-normality. It also has limited applicability when dealing with non-independent or non-normally distributed data.

In the final analysis, Cohen's d stands as an enduring testament to Cohen's vision of broadening the statistical paradigm to include effect sizes. It's a staple in modern research, offering a balanced, nuanced view of statistical results, and it's here to stay. As researchers, embracing tools like Cohen's d allows us to communicate our findings more effectively and intuitively.