Randomization of filter: days

[c-sturm]

Hi ESMIRA team!

I have a proposal: I know that it is possible to randomize the times in a day that questions can be sent out. It would be helpful to our current study if it would also be possible to randomize the days of the week that the questionnaires are asked. Do you think this is something that can be implemented?

Thanks so much, regards,
Corina

[SelinaDev]

Hi Corina,

thank you for the suggestion. I think that such a feature could indeed be useful.

However, unfortunately I think this might be harder to implement than it seems. The way things are now, ESMira randomizes signals for each day ahead of time. E.g., if you have a schedule with 5 randomized alarms, ESMira will figure out when each of those alarms happens before the window for those alarms even begins. If you are inside the window, for example because the window starts at 9 am and you join the study at 11am, then ESMira will not even schedule for that day.
This leads to two problems. First, we would have to figure out how to schedule alarms in a way, that you don't have to wait for half a week before you get any notifications. I think that's relatively easy to solve, but still more work than simply extending the current system. The second (bigger) problem is that iOS is rather limited in terms of the number of alarms that can be scheduled ahead. We've hat multiple people run into problems with that recently. So if we were trying to just schedule ahead for a whole week, that would mean more alarms to schedule, and users would run into that limit even sooner. I think this can also be solved, but only by using a radically different approach to scheduling than we currently have.

I will definitely put that suggestion into our backlog, but for these reasons I don't think that if I get to pick it up, it would be any time soon. However, in the upcoming ESMira release we will introduce programmatic filters. These will allow you to specify a script that ESMira will run to check if a questionnaire is currently active. We already have a day() function in our scripting system, which will give you the current study day (i.e., number of days the participant is already in the study). So if the randomization is not dependent on exact week days, that could be used to create a script to evenly space out the days a questionnaire is active on. And if a function to check the weekday would be required, that should be relatively straight forward to implement. The script filter functionality is relatively new, and I don't yet know how it will interact with the exact timings (e.g., if there is a false negative if a questionnaire would be inactive today, and ESMira tries to schedule for tomorrow), but I think it might be worth a try.

Best,
Selina

[c-sturm]

Thank you so much for the quick reply, then we will think of a different sampling plan in the meantime!
Regards,
Corina