Skip to content

Random Allocation

Distribute participants across experimental conditions

This is a method to enable a so-called between-subjects design. It allows you to generate a link that will lead participants to one of the experiments in your project, according to weights that you assign.

Setup

Create a separate experiment for each of your conditions, for example MyProject_A, MyProject_B (the naming is irrelevant). Save a version for each of these and create participant slots. Then, on your home page, under Distribute participants across experiments, press Create a new link. In this form, you'll pick a name for the new random allocation setup. Pick a name that makes sense for your project, and is unique. The name will become part of the resulting link to send to participants. On the next screen, you can add or remove experiments, and change their weighting. You can always make changes to this, by accessing it from your home page. The name can't be changed though (as it's used in the link). To disable an existing link, just remove all experiments, or set all weights to 0.

Weighting \& Slots

On the link form, you can increase and decrease the count for each experiment. These function as weights; the percentage column displays the relative weight of this experiment. For example, if you have three experiments with weights 1, 2 and 3, they will receive circa 1/6, 2/6, and 3/6 respectively, of the participants who follow the link. The total does not have to correspond to the total participants you wish to recruit, it's just about the proportion. You still have to make sure each of the experiments has a batch with open slots, or the participant will be told the experiment is not recruiting. Note that the allocation is not guaranteed to have that exact fraction, as it is random. But take into account that given drop-out rates for online experiments, it will be hard to achieve an exact distribution in any case.

Some experimenters recruit their participants from online recruitment platforms such as Prolific or Amazon Mechanical Turk. For this you'd typically need to pass a parameter related to the participant's identity to Meadows. In order to do this you'd select the appropriate template and incoming variables in the Batch form for each of your experiments. In this case, you have to make sure that you include these query parameters in the random allocation link as well.