We are conducting a 2 year longitudinal study with at least three time points of data collection. A nationally representative sample would be ideal, but attentiveness and retention are more important than nationally representative. My sense is that our best bet will be finding a way to recruit a national sample where we can get email addresses and manage our own sample.
Prolific and CloudResearch (mturk) have options for longitudinal studies, but with subject pool turnover, attrition rates are likely to be too high, and neither platform allows asking for email addresses to contact participants directly. Lucid allows asking for email addresses, but my experience with Lucid participants so far has been pretty bad--unusually high rates of failing attention checks.
How have others recruited and maintained national samples for longitudinal studies?
Article Understanding Group Dynamics: Theories, Practices, and Futur...
Read
In muy experiencia about this there two important issues to bear in mind. First, doing a kind of research interesting as well for the participants (the extreme of this would be the citizen science in which participant co-desing the research). And secod, provide individualized feedback to all the participants.
Cory Clark , to my understanding, for a sample to be statistically representative at national level, it has to be calculated and collected probabilistically, do Prolific and CloudResearch (mturk) allow for probabilistic sampling? (Depending on the topic of the research and the desired power, the required sample size could also be quite high).
Attrition is not a problem in a longitudinal (panel) design as it can be solved with a rotating panel by allowing for new-comers to compensate for the ones leaving the study, and later adjusting the (probabilistic) sampling weights mathematically to compensate for attrition. (See for example the rotational design of EU-SILC in the picture above.)
If the idea is to collect a non-probabilistic sample, then it will never be representative, but incentives could be used to keep the number of participants. Incentives of course should be managed with care, as it can bias the study (this is for example why monetary incentives tend to be avoided in some experimental designs).
Word-of-mouth has been an effective way to recruit participants but scalability depends on personal network size. Earned media (appearances in news stories, blogs, podcasts) can be a good way to recruit hundreds or even thousands of participants. However, this method is lumpy and time-consuming. We've tried paid social media with limited success -- many people sign up, but few engage deeply with the task (forecasting, in our case). Prolific and CloudResearch seem to do a good job but are expensive.
Overall, no method is perfect. Earned media is usually worth trying. But it often takes many tries to "earn" exposure that is useful for recruiting participants.
Hey Cory - you can recontact participants through Prolific. We're doing this currently. Can put you in touch with one of my students as to how.
- Badges
- Science topic
- Similar topics
- Medicine
- Epi
- Age Groups
- Children