Conversation

My friend

made a fantastic quiz based on a buttload of data where machine learning predicts what gender you are. I'm surprised it thought so strongly that I was male! programs.clearerthinking.org/gender_continu

258

Replying to

and

The quiz was pretty interesting. However, they claim that there were no large differences on the grounds that correlations was only 0.4 ish. But a correlation of 0.45 is equivalent to Cohen's d of 1.0, which is typically considered a large effect. doi.org/10.1177%2F2515

Replying to

and

There is a lot of debate over what a “large effect” is. According to this paper (journals.plos.org/plosbiology/ar) r=0.4 is roughly the median effect size for statistically significant results in psychology and cognitive neuroscience papers (d=.93)

journals.plos.org

Empirical assessment of published effect sizes and power in the recent cognitive neuroscience and...

Author summary Biomedical science, psychology, and many other fields may be suffering from a serious replication crisis. In order to gain insight into some factors behind this crisis, we have...

Replying to

and

fwiw intuitively in my surveys ive personally come to feel anything 0.4 and higher as 'strong', stuff around .3 as 'medium' and .2 and lower as 'mild'

9:31 PM · Oct 12, 2021

Replying to

and

The correlation between the closing over/under line and the actual total for NFL games is ~0.3 So, yea, assuming the relationship is linear-ish, 0.3 gives you plenty of information to work with