Millisecond Forums

Question about interpret IAT data

https://forums.millisecond.com/Topic24967.aspx

By zphermione - 5/15/2018

Hi everyone,

I am encountering a data analysis problem and hope you could help me!  I am doing a research about letting participants read three positive or negative articles about female politicians and measuring their IAT before and after reading articles. As I read posts before, the association score between the target and attribute is the expression.d score in the output data, so I compared the means of expression.d scores between the before and after experiment IAT test. The result shows that participants who read positive articles, expression.d changed from-0.25 to -0.12. Does it mean their attitude has been improved? Or it means nothing? Or I am on a wrong way to deal with the data? 

Thank you very much for your help!
By Dave - 5/15/2018

zphermione - Tuesday, May 15, 2018
Hi everyone,

I am encountering a data analysis problem and hope you could help me!  I am doing a research about letting participants read three positive or negative articles about female politicians and measuring their IAT before and after reading articles. As I read posts before, the association score between the target and attribute is the expression.d score in the output data, so I compared the means of expression.d scores between the before and after experiment IAT test. The result shows that participants who read positive articles, expression.d changed from-0.25 to -0.12. Does it mean their attitude has been improved? Or it means nothing? Or I am on a wrong way to deal with the data? 

Thank you very much for your help!

> The result shows that participants who read positive articles, expression.d changed from-0.25 to -0.12

This depends on how you coded your categories. Generally speaking, both -0.25 and -0.12 are in the same direction. Using the conventional breakpoints, -0.25 would indicate a "slight" preference, whereas -0.12 would essentially indicate no preference in either direction (it's below the breakpoint for a slight effect; cf. https://implicit.harvard.edu/implicit/demo/background/raceinfo.html ).

As for how to best deal with this kind of design and data, the best advice I can give you is to look at other studies that have employed IATs in pretest-posttest designs.