0
$\begingroup$

I conducted a two-way between groups ANOVA to test the impact of video gaming (gamer/non-gamer) and gender (male/female) on reaction time. I'm conducted my analysis using SPSS and received the following result: http://imgur.com/8EHKvxN

As you can see, there is no interaction and gender was not significant. However, being a gamer did have a significant impact on response times. I need some help with the effect size. I'm not sure if it's best to use Cohen's criteria (0.01 = small, 0.06 = medium and 0.138 = large) or a different effect size estimate. 0.234 seems ridiculously large in comparison to Cohen's criteria, does this seem right?

I really appreciate your time and help. Thanks.

Improve this question
$\endgroup$
  • $\begingroup$ How did you compute .234 and for what purpose ? why Anova? state your purpose? $\endgroup$ – Subhash C. Davar Jun 19 '16 at 15:21
3
$\begingroup$

Cohen's criteria for labeling effect sizes is abstract and in relationship to the variance rather than into the variable being studied.

Imagine if you will that the difference between means was the same as what you have now, but your variability in measurement was half what it was. Would you consider the effect to be 'larger' in the common sense of the word?

In short, don't get hung up on the ritual of calling an effect size large medium or small. Instead, be sure to (additionally) report the mean differences and SD of those differences. Trust your audience to judge for themselves the 'bigness' of the effect you are presenting.

Also, partial eta isn't Cohen's d.

Improve this answer
$\endgroup$
  • $\begingroup$ Thanks for the reply. Sorry if I wasn't clear. I know that partial eta squared and Cohen's d are separate effect size statistics, what I'd like to know is if I should simply use partial eta squared because that is the output I receive from SPSS, or if I should calculate a different one such as Cohen's d? Also, I'm referring to the SPSS Survival Manual by Julie Pallant which says that Cohen's criteria that I mentioned above can be used to interpret partial eta squared, and my result just seems rather high in comparison. Thanks again $\endgroup$ – Michael Jun 20 '16 at 8:12
  • $\begingroup$ The choice of an effect size metric is a mixture of tradition and prefered interpretation. Do you think your results should be about the proportion of variance explained? If so, use eta. Do you think your results should be about the magnitude of the difference relative to the variance? If so, use cohen's d. For my part, given the choice between the two and the type of study you describe, I'd lean towards reporting cohen's d. $\endgroup$ – russellpierce Jun 20 '16 at 13:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.