Randomization into control or intervention group during questionnaire

I am about to begin a pilot randomized control trial - and would like ODK to randomly assign new participants into a control or intervention group - with a max amount of participants in each group. Is there a way for this to be done automatically in ODK - while our study staff are completing the interviews. We are using Samsung tablets and ODK Central.

Hi Jordyn,
This is a great question.
This should be possible to a degree, but might need a bit more info from you.

For example, you could use a 'calculation' question to generate either a random 'control' or 'intervention' tag for the participant when you're registering them.

So, for example, say you are interviewing 100 people.

You want 40 people in your control and 60 in your intervention group, for example.

You can use a calculation to randomly generate a number, if the number falls under 40%, for example, then tag them 'control'. Otherwise, tag them 'intervention'. This means that the form bases the tagging on a percentage basis, not on an absolute maximum per group. It's random, so you'll get 40~ish in the control group and 60~ish in the intervention group.

If you want to ask them separate questions based on their 'tag', then you would just use have two sets of questions in there and use skip logic/relevant to apply certain questions to the control group, and other questions to the intervention group, for example.

Let me know if that makes sense, or if you need further guidance! Other people might have interesting ways of solving this too, but that's where my mind immediately jumped to!

I think, strictly as specified, the problem is a little ill-posed. Specifically, if there is a predetermined maximum group size, and you randomly assign a sequence of incoming elements to one of 2 groups, then soon as you happen to hit the max size of either group then all subsequent elements must therefore be strictly assigned to the other group; which isn't random [sic].

To accomplish a truly random assignment of folks to two groups then I believe you'll need all the individuals up front; is that possible? [and does what I said make any sense? :slight_smile: ]

1 Like

Greetings from field work in Laos! This makes perfect sense, and I ended up using the random number generator. Thank you for the excellent recommendation.

1 Like

Hi! thank you for your reply. this does make sense but unfortunately we don't have the list beforehand. I realized that being strict with the numbers in each group wasn't advised, as if the groups vary slightly in their numbers it wouldn't make a difference (statistically speaking). So I ended up trying the random number generator, and I will keep an eye on if this works well or not.

Well that is fantastic! Glad it worked for you - I love using 'random' for all sorts of fun little things...this is a great use case :slight_smile:
Good luck with the data collection - hope it all works out great!