Millisecond Forums

Hi,

I have thee questions and I would be extremely gratefull if someone could answer these!

question 1: 

I am programming an SC-IAT with a practisetrail at the very beginning and then 2 test trials (two seperate SC-IAT's - 2 brands).  There aren't any practisetrails before the test trails of which I want to use the data, so the D's I am using are the d1 (1st test) and d2 (2nd test), see below:

/ d1 = (m2a - m1a) / expressions.sda

/ d2 = (m2b - m1b) / expressions.sdb

Suprisingly one of my d1 was 1.04.  Just recently I found out that it is possible to have a d between -2 and +2, and I assume that this means that someone has an extremely strong positive attitude towards that category?!  Unfortunately the <% expressions.attitude %>  in the summary changed to the negative - so it said 'onaantrekkelijk' instead of 'aantrekkelijk' (unattractive/attractive) (script see below):

<trial summary>

/ stimulustimes = [0=summary]

/ validresponse = (" ")

/ ontrialbegin = [values.magnitude = "weinig of niet"]

/ ontrialbegin = [if( abs(expressions.d1) > 0.15 ) values.magnitude = "enigzins"]

/ ontrialbegin = [if( abs(expressions.d1) > 0.35 ) values.magnitude = "gemiddeld"]

/ ontrialbegin = [if( abs(expressions.d1) >= 0.65 ) values.magnitude = "heel"]

/ ontrialbegin = [if(expressions.d1 >= 0.0) expressions.attitude = "aantrekkelijk"]

/ ontrialbegin = [if(expressions.d1 < 0.0) expressions.attitude = "onaantrekkelijk"]

</trial>

<text summary>

/ items = ("Bedankt voor je deelname! ~n~n

Je <% item.targetALabel.1 %> Brand Appael score was <% expressions.d1 %>, wat betekent dat je <% item.targetALabel.1 %> <% values.magnitude %> <% expressions.attitude %> vindt.~n~n

Je <% item.targetA2Label.1 %> Brand Appael score was <% expressions.d2 %>, wat betekent dat je <% item.targetA2Label.1 %> <% values.magnitude %> <% expressions.attitude %> vindt.~n~n~n~nDruk op de SPATIEBALK om de taak af te ronden.")

/ size = (60%, 60%)

/ fontstyle = ("Arial", 3%, true, false, false, false, 5, 0)

/ color = (65, 188, 214)

/ hjustify = left

</text>

To say 'aantrekkelijk' I assume that the d1 has be >= 0 and not untill 1.  Could someone maybe help me understand what happend? And what the chances are that a D can be above 1, since then it might be a good idea to incorparte something like D >= 1. values.magnitude = extreem.

question 2: As I understand the improved algorith incorparates error penalties.  I assume this means someones D is lower if there responces have a lot of errors? Could this error penalty also be incorparted in the calculation of the D for the data file? without having to use SPSS?

question 3: Now I can't see the D's in the data file, to see them do I only have to remove '/ recorddata = false' from the summary?

script included

The D computed from m2a and m1a is the D computed from practice trials (see: http://www.millisecond.com/forums/Topic3444.aspx). Since you don't have any practice trials, I find it very likely that you simply broke the scoring in the process of modifying the original script. See this thread for a more detailed explanation of the different means involved in computing D: http://www.millisecond.com/community/forums/t/1370.aspx.

Again, since you probably broke the scoring, I wouldn't trust this value in the first place. However, it is correct that D varies between -2 and +2 and as such a value of 1.04 is theoretically possible (but should be rare in practice). Whether the respective value reflects a strong positive or a strong negative association depends on how you've set up the different categories in your IAT. The first thread I referenced has more info on that, however...

... suggests that (a) you may have entered categories in the wrong way and/or (b) you  broke the scoring.

No, see my point above.

D scores above 1 should be relatively rare in practice, however this may differ for some domains. You should check the relevant literature, especially look for available meta-analyses. However, it is way more likely that you broke the scoring (yes, I know I'm repeating myself) and the D scores you're getting actually don't mean anything. That said, I don't think that introducing another scoring quantifier ("extreme") is warranted, but go ahead and do whatever you want.

Yes.

No. If a participant makes the same amount of errors in both pairings (compatible and incompatible), D will be basically unaffected. If a participant makes systematic errors (e.g. only in the incompatible pairing), the resulting D will be larger.

All the IAT scripts available from millisecond.com adhere to the conventions of Greenwald et al.'s "improved" scoring algorithm, i.e. error penalties are included. However, you probably broke the scoring.

Because you probably broke the scoring. Removing the 'recorddata' attribute won't help here, but feel free to try nevertheless.

No -- you *always* post random data files for some reason.

~Dave

Hi Dave,

Thank you for your quick reply.  I know that the D computed from m2a and m1a is the D computed from practice trials, but in the <block> elements of the incompatible, compatible tests 1 & 2, I used 'a' for the 1st test (instead of for the practice trial) and 'b' for the 2nd test.  Seeing that I don't have 2 sets of results (practise and test) I don't have to devide the results by 2, so I thought I could simply delete the 'd' in the original script and use da (test1, now d1) and db (test2, now d2).  Is the D computed from m2a and m1a then restricted to practicetrials?  How could you then report two separate D's? (with two single tests (incl. compatible/incompatible ofcourse?)

Hereby the script instead of the data file - my appologies!

thank you, Celeste

Hi Celeste,

just writing to let you know that I'm not going to look into the script you posted -- at least not anytime soon. Reasons: (1) I really don't have the time to reverse-engineer everyone's (potentially flawed) programs, 'cause I've got other stuff to do and (2) you should have all the required info* at your disposal to check / figure this out by yourself. Plus, you'll learn more if you dig into this on your own instead of having other people fix it for you. Perhaps though some other user feels inclined to check into this. Good luck!

Best wishes from a fellow Inquisit user,

~Dave

* That would include a solid understanding of how D is computed and of the Inquisit IAT scripts.

Dave, I totally understand you don't have the time to look through the entire script. 

But could you reply if you can use the 'a' (typically used for practise test) for a 1st test and generate one D? Because if that is possible, I can not find a mistake in my script, seeing that the 'a' now refers to the 1st test and the 'b' to the 2nd.  This therefore generates two seperate d's - d1 and d2.

Could any one else have a quick look for me please otherwise?

Cheers,

Celeste