Millisecond Forums

Randomization of items

https://forums.millisecond.com/Topic30902.aspx

By kveylet - 2/22/2021

Hello, I apologize if this is already answered best somewhere else, but I was wondering if there is any way to control the randomization of selecting items between different participants. I ran a study in which people heard twelve different kinds of sounds (music theory intervals, e.g. a major second or perfect fifth), and for each type of sound could randomly hear one of three versions of it.
MWE with just one interval type, in case that is useful:
<item min2>    / 1 = "a1.16-15.mp3"/ 2 = "a2.16-15.mp3"/ 3 = "a3.16-15.mp3"</item>

<video min2>    / items = item.min2        / loop = true    </video>

<block LDTaskMIN2>
/ bgstim = (wordlabel, nonwordlabel, video.min2)
/ trials = [1-52=noreplace(ldtword, ldtnonword);
            53=surveypage.MFR]
/ ontrialend = [
            values.PitchInt = video.min2.currentitem;
            if(values.PitchInt == "a1.16-15.mp3") values.OctLevel = "1"
                else if (values.PitchInt == "a2.16-15.mp3") values.OctLevel = "2"
                else if (values.PitchInt == "a3.16-15.mp3") values.OctLevel = "3";
            values.IntClass = "min2"
]            
</block>


In analyzing responses from complete submissions, there were some sounds with an equal amount of participants across each of the three versions, and others with a fair amount of disparity. Could this be driven by participants who launched the task but did not finish? Is there any way to attempt to coerce equal selection of the three varieties of an item?

Thank you so much for your time!
By Dave - 2/22/2021

kveylet - 2/22/2021
Hello, I apologize if this is already answered best somewhere else, but I was wondering if there is any way to control the randomization of selecting items between different participants. I ran a study in which people heard twelve different kinds of sounds (music theory intervals, e.g. a major second or perfect fifth), and for each type of sound could randomly hear one of three versions of it.
MWE with just one interval type, in case that is useful:
<item min2>    / 1 = "a1.16-15.mp3"/ 2 = "a2.16-15.mp3"/ 3 = "a3.16-15.mp3"</item>

<video min2>    / items = item.min2        / loop = true    </video>

<block LDTaskMIN2>
/ bgstim = (wordlabel, nonwordlabel, video.min2)
/ trials = [1-52=noreplace(ldtword, ldtnonword);
            53=surveypage.MFR]
/ ontrialend = [
            values.PitchInt = video.min2.currentitem;
            if(values.PitchInt == "a1.16-15.mp3") values.OctLevel = "1"
                else if (values.PitchInt == "a2.16-15.mp3") values.OctLevel = "2"
                else if (values.PitchInt == "a3.16-15.mp3") values.OctLevel = "3";
            values.IntClass = "min2"
]            
</block>


In analyzing responses from complete submissions, there were some sounds with an equal amount of participants across each of the three versions, and others with a fair amount of disparity. Could this be driven by participants who launched the task but did not finish? Is there any way to attempt to coerce equal selection of the three varieties of an item?

Thank you so much for your time!

You're selecting items at random, and Inquisit running on participant A's device has no way whatsoever of knowing whatever happened or happens on participant B's device. So no, if you're going to select items like this, there is no way to control randomization across participants.

What you would have to do, perhaps, is set up between-subjects conditions such that an equal amount of each sound's versions is administered across the X conditions, then assign an approx. equal amount of participants to each condition