Millisecond Forums

Thanks for the tips Dave. :)

+x

 

Sercan - 5/12/2026

We have set up our online study such that our 8 experimental conditions are block-randomized by participant number, i.e. the first 8 participant numbers are assigned to one of the 8 conditions, so are the next 8, etc.; the intention was to ensure all 8 conditions get an equal number of participants.

Though the code works as intended, the block randomization has failed because the majority of people who visit our experiment page, never click 'Start'; Despite that, they are assigned a participant number (as far as I can tell), so we end up with many participant numbers that never did the study. This has resulted in our assignment of conditions being, in practice, fully random, so some conditions have received far more participants than others.

For future studies, do you see a way by which we can design our inquisit experiment code to achieve what we wanted here?

There is nothing you can do against pre-launch dropouts, and as you correctly assumed, the critical inputs (subject ID, group ID) are and have to be assigned before launch, when a participant loads the experiment's launch page.

If there's a better way depends on how you recruit participants. If you have just a public link posted somewhere or in multiple places, which an unknown number of people may or may not click, then there really isn't much you can do. If you have more control over your participant pool, you might pass the critical input(s) in via URL parameter instead of having them generated randomly by the launch page. Suppose you have a pool of 100 participants and four conditions assigned via group ID. Then you would give 25 partcipants a link to run condition one,



another set of 25 for condition two,



and so forth.

Any dropouts you can compensate for by recruting additional participants to fill up the respective conditions.

Or you can do a first round with an uncontrolled, public sample as you did in the current case, and then do a 2nd round of targeted recruitment as described above to bring cells up to approximately equal numbers.

We have set up our online study such that our 8 experimental conditions are block-randomized by participant number, i.e. the first 8 participant numbers are assigned to one of the 8 conditions, so are the next 8, etc.; the intention was to ensure all 8 conditions get an equal number of participants.

Though the code works as intended, the block randomization has failed because the majority of people who visit our experiment page, never click 'Start'; Despite that, they are assigned a participant number (as far as I can tell), so we end up with many participant numbers that never did the study. This has resulted in our assignment of conditions being, in practice, fully random, so some conditions have received far more participants than others.

For future studies, do you see a way by which we can design our inquisit experiment code to achieve what we wanted here?