How exactly does Inquisit assign subject ids in web experiments?


How exactly does Inquisit assign subject ids in web experiments?

Author
Message
AKrishna
AKrishna
Associate Member
Associate Member (230 reputation)Associate Member (230 reputation)Associate Member (230 reputation)Associate Member (230 reputation)Associate Member (230 reputation)Associate Member (230 reputation)Associate Member (230 reputation)Associate Member (230 reputation)Associate Member (230 reputation)

Group: Forum Members
Posts: 11, Visits: 15
Hi all,

I'm sifting through the data from a recent online experiment and I noticed something odd. My experiment was run using a batched file which started with a consent block followed by a demographics block, looking like this:
<batch>
/ subjects=(1 of 1)
/ groupassignment = subjectnumber
/ file="consent.iqx"
/ file="Demographics.iqx"
/ file="MainExperiment.iqx"
</batch>

Subject number was set to be assigned in ascending order. Now, the first question in the demographics block was age and it was gated to accept only responses within a certain range (using /mask). As we had mentioned the age limitation in our online recruiting, we expected participants outside that range to quit. Although I had posted that participants could quit the experiment using Ctrl + Q at any time, some participants who were outside of the desired age range had forgotten or skipped those instructions and so quit the experiment using task manager or by rebooting.

I would have expected the subject ids assigned to these people to be associated with no data or maybe just a line with the unacceptable response, which was the case in the data file generated by the demographics block. However, the data file for the main experiment also contained full datasets associated with these ids. Does anyone have any ideas why this might be? To me, it looks almost as though Inquisit assigned the subject number twice, with the initial assignment producing the data for the demographics block and the second assignment producing no data in that block, but otherwise producing data as normal in the main experiment. Is this possible? If so, is there any way to recover the demographic data for the second assignment of the subject id? Unfortunately, this is a fairly serious issue, as I cannot identify the Prolific codes of those second assignments and therefore cannot pay them. Would appreciate any advice! I can post code and datasets if needed.

Best,
AK

AKrishna
AKrishna
Associate Member
Associate Member (230 reputation)Associate Member (230 reputation)Associate Member (230 reputation)Associate Member (230 reputation)Associate Member (230 reputation)Associate Member (230 reputation)Associate Member (230 reputation)Associate Member (230 reputation)Associate Member (230 reputation)

Group: Forum Members
Posts: 11, Visits: 15
I was able to find out what had happened with some belated feedback from participants. It is expected behavior for Inquisit to start the next file in a batch if the current file is quit using Ctrl+Q. My participants simply continued the experiment without entering further data. The reason I was worried is because I would have expected to read "Ctrl+Q" as a response in the data, but instead saw only the number participants had entered in the age textbox. So for anyone experiencing similar issues or worries, you need to be aware that pages structured like this:

<surveypage info_1>
/questions = [1=age]
/stimulustimes = [0 = mousegone]
/ showpagenumbers = false
/showquestionnumbers = false
</surveypage>

<textbox age>
/mask = positiveinteger
/ caption = "What is your age:"
/range = (18,50)
</textbox>

<text mousegone>
/items = ("On some systems, the mouse cursor may disappear on this screen.
If it does not reappear when you move the mouse, you can use the Tab key to
select the Finish button and then press Enter to continue.")
/ fontstyle = ("Arial", 1%)
/position = (22%,24%)
/halign = center
/txbgcolor = black
/txcolor = white
</text>


...do not record responses of "Ctrl+Q" in the data file in the way that most <trial> elements do. They can in fact even record invalid responses instead (for example, my data file showed that one participant who presumably quit had responded "51" to the age question, which was an invalid response due to the /range attribute). An interesting quirk of the software I didn't know about, hopefully this information may help other people with related questions!

GO


Reading This Topic


Login
Existing Account
Email Address:


Password:


Social Logins

Select a Forum....






Millisecond Forums


Search