By honglingggs - 2/4/2015
Hi,
I am currently conducting a test (GO/No-GO Association Task) to examine the implicit attitude towards individuals with disabilities. I understand that typical GNAT requires the derivation of d'prime indices for each critical block in order to compare the association strengths between targets and attributes.
However, I would also like to quantify a single index for implicit attitude in order to run subsequent analyses looking at its relationship with explicit measure of attitude. I am uncertain of the best way to do so though. Should I run the analyses based on the difference score by subtracting d'prime(disabilties+plesant) from d'rime(disabilities+unpleasant)? Or would it be more desirable to conduct the analyses with either the d'prime of (disabilities+pleasant) or (disabilities+unpleasant) alone?
I am more inclined towards the latter idea because I feel that the d'prime for each block is distinct on its own as weaker association between disabilities+pleasant may not necessary translate to a stronger association between disabilities+unpleasant. I feel that both are on relatively different scale. However, I have seen literature which have collapse both d'primes by deriving a difference score to quantify an implicit attitude index.
Anyone can advice? Thank you in advance!
|
By Dave - 2/4/2015
My quick thoughts: d' is a dimensionless statistic, so -- by definition -- there is no issue of scale. (To be strict: d' does not represent association in the sense of, say, a correlation coefficient. d' reflects the separation between the means of the signal and noise distributions under assumptions of normality.)
|
By honglingggs - 2/4/2015
Hi Dave,
Thanks for your reply. What would you suggest then if I am interested to explore the relationship between the implicit measure and the explicit measure?
I would definitely need to quantify an index for the implicit measure in this case and D score (for IAT) might not be ideal as GNAT involves response deadline.
Look forward to your reply.
|
By Dave - 2/4/2015
I see no problem with computing a difference score. However, this is a methodological question completely unrelated to Inquisit, so there might be better venues for you to discuss this (e.g. mailing lists or fora dedicated to statistics and research methods).
|