Let's say we are doing a random sample of a university to determine how many of the students participated in some sort of volunteering in the past year. You randomly select 2000 students and receive a point estimate of hours spent volunteering. In this scenario, those that have completed some sort of service are more likely to complete the survey. Is there any literature that would tell you how far to extrapolate that data, i.e., how you might adjust or weight the data to be more accurate based on the potential response bias?
Hi Nathan,
this is an extremely tricky topic , since you have to cope with social desirability. I do not think that there is an easy answer to that, but I guess that in the first place you have to cope with a self-selection bias, as you already wrote, and even those students who respond might tend to exaggerate.
In theory, the best way to come up with weights would be to have a control group of which you exactly know how much they volunteered and then you compare their answers with the real workload. However, this might be really cumbersome to do.
Here is some excellent literature which is closely related to the topic:
https://www.semanticscholar.org/paper/Survey-responses-%3A-an-evaluation-of-their-validity-Wentland-Smith/fb3592cc10791bc1e913e759cda4de994280db2e
(Wentland and Smith, 1993, Survey Responses: An Evaluation of their Validity, Academic Press)
The authors provide an overview of numerous studies which used triangulation (i.e., different kinds of measurement in parallel). In a nutshell, the results show that people lie. The amount depends on the sensitivity of the subject. If you ask people about their income, for example, almost everybody will lie...
I guess that you find related studies in the book which give you an idea what weights you can use.
Hi
There is a relationship between the research community and the size of the research sample. You can use statistical books to determine the size of your sample
Your question is not clear, I don't understand what you mean, but in general you must make normal distribution test, in order to determine the type of your data and then you can use the suitable statistical analysis.
- Badges
- Science topic
- Similar topics
- Higher Education
- Universities