Millisecond Forums

AAT scores corrected for accuracy

https://forums.millisecond.com/Topic15483.aspx

By JRS - 2/5/2015

Hi Dave
I trying to wrap my head around the problem of getting two kinds of recorded latencies from my AAT script
I'm using the version from the library which has expressions to compute difference scores based on all push-pull responses (which work fine and I want to use)

e.g.
<expressions >
/ AAT_Diffscore_Cat1 = if (values.expcondition == 1) trial.AAT_1.medianlatency - trial.AAT_2.medianlatency else
trial.AAT_2.medianlatency - trial.AAT_1.medianlatency
.
.
</expressions>

The problem is that I want to get the same eventual median difference score(s) for latencies corrected for accuracy. I think there's something that I need to modify in the following section of the script but I don't know what or which variables to start with.

For example in

<trial AAT_1>
.
.
/ ontrialend = [if (trial.AAT_1.correct) values.sumRTcorrect_1 += trial.AAT_1.latency]
/ ontrialend = [values.sumRT_1 += trial.AAT_1.latency]

/ontrialend = [if (trial.AAT_1.responsey < values.mouse_y) values.response = 1 else values.response = 2]
/ ontrialend = [values.mouse_change = abs(values.mouse_y - trial.AAT_1.responsey)]
/ ontrialend = [values.mouse_y = trial.AAT_1.responsey]

/ ontrialend = [values.RT = trial.AAT_1.latency]
/ ontrialend = [values.correct = trial.AAT_1.correct]
.
.
On the other hand, maybe the solution is not to modify that part of the script because latencies corrected for accuracy can be computed at <expressions> using the existing variables?

any help and advice is appreciated

kind regards

Jonathan
By Dave - 2/5/2015

That expression relies on the *median* latency, not a simple mean (which is easy to compute "by hand" via values, etc.: sum up the relevant latencies and divide by number of relevant trials). If you want to calculate median RTs by hand, you have to jump through additional hoops:

https://www.millisecond.com/forums/Topic14637.aspx

It's not as easy as just using a bunch of existing values. It may be preferable to leave the script as is and do this later based on the raw data using your preferred statistical analysis application.