Explore BrainMass

Effect Size

The effect size is a measure which quantifies the magnitude of the relationship between the variables being compared in a statistical analysis. The computation of the effect size is important for understanding whether or not any meaning or importance is associated with a statistical difference. If a researcher is able to reject the null hypothesis, all this means is that a statistical difference exists, but not that this difference is important.

Effect sizes can be easily calculated and are standardized to a common scale which allows for comparisons to be made between different effect size derivations1. The general formula to compute the effect size is as follows2:

Formula: (mean of the experimental group – mean of the control group)/(standard deviation of the control group scores)

The effect size can take on a negative, neutral or positive meaning. Firstly, effect sizes are expressed as decimal numbers, with larger values being indicative of a higher level of importance being associated with the difference found2. Generally a value greater than 0.5 implies that the difference found is important, but as a general rule, a value of 0.8 or greater is indicative of a powerful treatment1, 2.

A value of 0.00 signifies that a neutral relationship exists between the control treatment and the experimental treatment. Thus, no difference is evident. A negative value states that the control group performed better than the experimental treatment1.

Evidently, the effect size is a rather critical concept in statistics because it is crucial to comprehend whether the results a researcher collects have applicable importance to real-life situations. Without knowing the effect size, the information collected from hypothesis testing is purely statistical in relevance.




1. My Environmental Education Evaluation Resource Assistant. (2014). Power Analysis, Statistical Significance, & Effect Size. Retrieved from:

2. Villanova University. (2014). The Concepts of Statistical Power and Effect Size. Retrieved from:

Image Credit: Wikimedia Commons

Repeated-Measures Design and T-tests

John wants to measure self-esteem among victims of domestic violence before and after they complete an 8-week program for abuse victims. Could he use a repeated-measures design and t-test for this study? Why? Give John directions for how to set up his study as a repeated measures design. Be specific about who, what, when, and ho

Outlining a Possible Study with Independent/Dependednt Variables

Fill in the following for a possible study with one independent variable (IV) with two conditions/treatments and a dependent variable (DV) that is measured on a continuous scale (interval or ratio): Independent variable = ______________ Condition A = ______________ Condition B = _________

Statistics: inference for proportions

Geogrge and Jerry are working on their final project for AP Statistics. They wish to investigate the question, "Do people really read documents they are asked to sign?" They believe that fewer than half of all people read a document they have been asked to sign. To test their theory, they randomly select 50 individuals rom va

Testing for the Level of Risk

Two rival book companies offer a special discount deal to a college for textbooks in math and science. Since the discount is the same and the quality of books appears to be the same, the manager of the book store has to make a decision. She conducts a random sampling of prices of math and science books from both companies. Be

Statistical Power - Size

Chapter 6 talks about effect size and statistical power. What are these? How are they relevant to statistics and how do you use them? Since it is talking about effect size do you always want big or do sometimes you want a small or medium? How do you determine this? You can discuss and explain these topics by explaining what a

Mini meta analysis

Conduct a mini-meta-analysis. There is really no such thing as a mini-meta-analysis, but you don't have time to conduct a full meta-analysis. Anyway, with the topic of your choice, find 1-3 articles (WWW or Cybrary). Using information from the 1-3 articles, estimate the effect size of a variable. So, if you find more than one ar