# Interpreting effect size

One of the most widely used interpretation of effect size is Cohen's $d$, as follows:

• $<.10$: trivial
• $[.10,.30]$: small to medium
• $[.30,.50]$: medium to large
• $>.50$: large to very large

This appears to be the standard in psychology.

I am wondering if there is such a standard in sociology or social science in general?

I know the size of effect is best interpreted in the context of the research!

• If you already know that these "standard" effect sizes are a poor idea and best interpreted within the context of particular research why are you asking the question? – John Feb 28 '12 at 3:11
• I never said it is a "poor" standard! Statistical analysis is as much about interpretation (and commonsense) as it is about calculation. – Amarald Feb 28 '12 at 4:43
• IMO this interpretation is based on the fact that it is darn hard to find any large effect sizes in social science, whereas in other disciplines may run the rule that the threshold for large is 0.9. – Michelle Feb 28 '12 at 5:10
• But even within a field it's dangerous to have such a standard. You can only do it within a particular domain of research. Within psychology itself the standard should be vastly different for fMRI results, the Stroop effect, occulomotor capture, and weapon focus on eye witness testimony. Furthermore, whether it's in a field situation or laboratory situation should also influence the standard. – John Feb 28 '12 at 16:13

These "standards" in psychology are an unfortunate consequence of poor statistics training. Don't look for such standards in an entire field. At best they could be found within a particular subject matter. Cohen never intended these to be standards and just suggested them as a starting point for interpretation based on a prior analysis of social science effect sizes and intuition. We were supposed to grow beyond that suggestion, not turn it into doctrine.

John has already given a spot on answer. Just as an addendum (which is just a bit too long as a comment), let me just add a quote from Cohen himself (from Statistical power analysis for the behavioral sciences, 1988):

The terms "small,", "medium,", and "large" are relative, not only to each other, but to the area of behavioral science or even more particularly to the specific content and research method being employed in any given investigation [...]. In the face of this relativity, there is a certain risk inherent in offering conventional operational definitions for these terms [...]. This risk is nevertheless accepted in the belief that more is to be gained than lost by supplying a common conventional frame of reference which is recommended for use only when no better basis for estimating the ES index is available. (p. 25).

The emphasis is mine. In many cases, there is a better basis, since effects are often measured with scales for which we have some prior knowledge/intuition about the meaning of the raw units and the amount of variability in the outcome.

In his famous 1994 paper ("The earth is round (p < .05)"), Cohen himself also recommended moving away from 'standardized' measures of effect and instead advocated working with raw measures of effect. Another quote:

To work constructively with "raw" regression coefficients and confidence intervals, psychologists have to start respecting the units they work with, or develop measurement units they can respect enough so that researchers in a given field or subfield can agree to use them. In this way, there can be hope that researchers' knowledge can be cumulative. (p. 1001).

Again, emphasis is mine. It is unfortunate that Cohen got his name attached to these 'canned' values, when in fact he was quite careful not to overemphasize their meaning.