I am preparing my longitudinal study on the process of forming the vision of own adult life in emerging adulthood. I want to conduct two measurements (8-12 month break) and finally have (in a second measurement) about 100 people (50% men and 50% women). In each measurement participants will have to answer on one open-question (about vision of their own future) and fill out 3-4 questionnaires (summary time of whole one-time examination: c.a. 40 mins).
Basing on your experience - how many people should I test in the first measurement, to have at least 100 participants in the second measurement?
Thank you for your support!
Kamil Janowicz
Hi Kamil
power can be easily computed by such software as gpower (please google for it).
More important is, whether dropouts will create a bias (regardless of power). The keywords are attrition, differential dropout and endogeneous selection:
Elwert, F., & Winship, C. (2014). Endogenous selection bias: The problem of conditioning on a collider variable. Annual review of sociology, 40, 31-53.
Article Endogenous Selection Bias: The Problem of Conditioning on a ...
Haneuse, S. (2016). Distinguishing selection bias and confounding bias in comparative effectiveness research. Medical care, 54(4), e23.
https://scholar.google.com/scholar?output=instlink&q=info:dc0Ta-4a4UsJ:scholar.google.com/&hl=de&as_sdt=0,5&scillfp=10304052588018631723&oi=lle
HTH
Holger
I don't know if there is a literature on dropout rates, but you might look at Dillman et al, Internet, Mail, and Mixed-Mode Surveys: The Tailored Design Method, for a system that maximizes initial response rates.
Hi David,
I provided two papers that discuss the issue. Please note that Dillman's method aims at increasing the response rate in the first place whereas dropout is when specific individuals leave the study after the treatment and before measuring its effect. It's a kind of non-random missing.
Best
Holger
The question here is about minimizing a drop out rate, and I don't know if there is a literature on that.
Of course, assessing bias in retention is important, but it doesn't address the initial issue of how to minimize drop outs.
We cannot predict or expect how many drop-outs would happen. That is the reason why we need to do a power analysis before conducting the study. The power analysis suggests us a minimum requirement (sample size) to ensure statistical staility of the study findings. So once you get the minimum requirement (sample size), it would be better for you to invite more participants than the minimum requirement. And please remember that ensuring "equipose" is important in CER (comparative effectiveness research): Article Comparative Effectiveness Research and Its Application to Nu...
Thank you everybody for your answers. In my questions I have mentioned about your experience with usual rate of drop-outs in questionnaires longitudinal studies. For me is important to know, more or less, how many (in%) more participants I need to get proper sample in a second measurement.
Dear David L Morgan
I agree and going with Dillman can be a way. But as I remember Dillman is a way to motivate passive non-responders (i.e. people who intend to participate but do not initiate the behavior). I don't know if this can be helpful for people who are already part of the study. But you are right. Efforts to increase motivation is valuable in any case:
Bosnjak, M., Metzger, G., & Gräf, L. (2010). Understanding the willingness to participate in mobile surveys: exploring the role of utilitarian, affective, hedonic, social, self-expressive, and trust-related factors. Social Science Computer Review, 28(3), 350-370.
Beyond that however, I found the Haneuse paper which I posted extremely helpful - not only for reducing bias afterwards (which is not possible if you don't have a model about the potential dropout-increasing variables)--it helps to speculate about potential factors. If you think of them prior to the study, you can include and control them which reduces bias of the dropout.
Therefore, my conclusion from our exchange would be to do both: Increase motivation, reduce barriers, and think about dropout driving factors (and include them in the survey).
All the best,
Holger
I understand your question but it is difficult to give a number for "usual" rates of dropouts. In general it is also not only drop-outs that has to be taken into account when performing a longitudinal study when you are concerned with final number of participants. Such thing as death is also something to take into account, but might not be as important for your study.
In the longitudinal yearly follow-up study of patients afflicted by stroke where we followed patients for five years, the drop-out rate was less than 10 % for each year but the total share of persons dying during follow-up was about 45 %. Median age for the population was 75 years so in cases like this you have take mortality into account too. However I have understood that a drop-out of less than 10 % per year is considered very low. In the litterature you can also read that a respons rate for questionnairs that is above 75 % is considered to be good. Initial respons rate is of course also of importance for the number of participant that is available for follow-up.
Åsa Rejnö , thank you for your answer. That is what is interesting for me. As I know I see that in many studies this dropout rate is different so I was looking for someone with experience in my field. Theese are young people (18/25 years old) - pupils in high school and students so I hope that % of dropouts will not be as high as in studies on people about 75 years.
Accioding to all answers, I think I will try to have about 30-40% more people in first measurement.
- Badges
- Science topic
- Similar topics
- Psychology
- Developmental Psychology