Blog of Sara Jakša

Notes about Statistics in Social Sciences

There are just two notes, that I think are interesting and they are relevant to the social sciences.

The first one comes from the Many Analysts, One Data Set article. This article gave multiple teams the same data and hypotesis to research. There were still a lot of differences in aproaches the teams took, the variables that they used from the dataset and the effect that they come. So they then recomend, that single analysist would need to make as many analysis as possible, and calculate how many of them would need to show an effect, to be confident that there is an effect. Even better would be, if analysis in science would be done in a crowdsourcing matter.

The second one comes from the One Hundred Years of Social Psychology Quantitatively Described. This used all the possible researches, that they could find, to try and see what kind of effect is expcted for each variable. This can be useful when testing hypotesis when using Basiyan statistics or for (at least) some calculations fo power and needed number of participants for that power. The table of effects is below.

Effect Number of Meta Studies Effect Size Standard Deviation
Agression 31 0.24 0.20
Attitudes 32 0.27 0.14
Attribution 36 0.14 0.14
Expectancy effects 16 0.16 0.22
Gender roles 19 0.18 0.13
Group processes 27 0.32 0.15
Health psychology 22 0.17 0.13
Helping behavior 14 0.18 0.16
Intergroup relations 28 0.19 0.18
Law 25 0.17 0.08
Leadership 42 0.25 0.18
Methodology 29 0.21 0.10
Motivation 12 0.15 0.12
Nonverbal communication 29 0.22 0.17
Personality 32 0.21 0.14
Relationships 32 0.22 0.12
Social cognition 22 0.20 0.19
Social influence 26 0.13 0.18
Total 474 0.21 0.15

I am quite interested in the personal differences, so I also copied the who things, where they went a bit more in the peronality difference. The first is the personality vs. situation problem (disclamer: they are about the same):

Effect Number of Studies Effect Size
Personality 16,282 0.19
Situation 17,631 0.22

And then for the sex differences (sans cognitive differences, since these are only variables from social psychology):

Effect Number of Studies Effect Size
Sex differences 83 0.12
Sex targets 0.08
Sex actors 0.13