Millisecond Forums

IAT Data Interpretation

https://forums.millisecond.com/Topic19646.aspx

By Juanp - 9/4/2016

Hello,

I wanted to ensure that I was interpreting my IAT data correctly.

Please correct me if I am wrong, but from my understanding, the D score that we attain from the IAT summary file can indicate how each participant “performed.”
That is, a positive score indicates a higher sorting/association between Target A with attribute A and target B with attribute B. Where as, a negative score would be the opposite.
For example, if my Target A label was Latino and my attribute(s) A were bad, a positive score would indicate faster sorting of Latino and Bad and European and Good (B label and Attribute b).

Given that D scores can range from -2 to 2, would a higher D score indicate a faster sorting/association between Target and Attribute?

Thank you!
Juan
By Dave - 9/5/2016

> That is, a positive score indicates a higher sorting/association between Target A with attribute A and target B with attribute B. Where as, a
> negative score would be the opposite.

Yes, as detailed (among others) at https://www.millisecond.com/forums/Topic3444.aspx

> Given that D scores can range from -2 to 2, would a higher D score indicate a faster sorting/association between Target and Attribute?

A higher D-score indicates a *greater difference* in mean response time between the "compatible" and "incompatible" condition. It does not necessarily mean "faster". A D-score is -- in essence -- the difference between

mean response time in *incompatible* trials
and
mean response time in *compatible* trials

expressed in standard deviations.

I.e., (mRTi - mRTc)/sd

Suppose you have the following two hypothetical subjects:

Subject #1:
meanRT(incompatible) = 1800ms
meanRT(compatible) = 1200ms
SD(overall) = 400ms

D(s1) = (1800ms-1200ms)/400ms = 600ms/400ms = 1.5

Subject #2:
meanRT(i) = 3600ms
meanRT(c) = 2400ms
SD(o) = 800ms

D(s2) = (3600ms-2400ms)/800 = 1200ms/800ms = 1.5

You'll notice that both have identical D-scores *despite* substantially different RTs. Hypothetical subject #2 on average takes twice as long to respond compared to hypothetical subject #1. So: Who is faster? Faster with respect to what? That's not the information D is supposed to convey.

Also see http://www.jakewestfall.org/misc/D_bound.html