An quantitative e-survey resulted in 28 respondents out of 400 responding. Can the researcher make analysis with just the 28? Will the out come be reliable?
The biggest problem with having only 28 respondents is not the number, but the small response rate (7%). This has the potential to produce an enormous non-response bias. If the 28 respondents were randomly sampled, the statistical process would address this through probability. What can you do about this? You could try to determine if there is non-response bias. This could be done in several ways. If you sent our reminders after periods of time, you could do a wave analysis comparing the results by waves (after each reminder). If the results are different, it suggests non-response bias. If you did not send out reminders, you could re-sample the non-respondents and compare the analysis of these with the original. Again, differences would suggest a non-response bias. Good luck, Jim
As James E. McLean points out, for most purposes the 7% response rate means you cannot do much with the data. But you do not provide much information. If your sample (the 28) were randomly allocated into two conditions and performed different tasks, then it would depend on the power. In some area, say psychophysics, 14 participants in each condition could be enough is the measurements were precise enough.
Along with the problem of bias, there is also the limitation due to the large standard errors associated with such a small sample. In technical terms, your statistical tests will lack "power," which means that it will be very difficult to produce significant results.
Ronit - you could perform simple descriptive frequencies i.e. percentages, histograms etc - but you would not consider reliability with such numbers. The results are purely descriptive, non-generalisable and, for some, runs the risk of not being that compelling.
It is not just the issue of small n as James E. McLean mentioned. One way to look at your data is to think of ways in which your respondents might be systematically different from the ones who did not respond - leading to bias in your sample.
Kind regards,
Usman
Dear Ronit Akomeah , if you managed to get the attention of 400 but only 28 responded, then one would immediately question the structure of your survey. Generalizability is an important aspect of quantitative research. It is usually backed up by a healthy sample size. Hence, proceeding with a sample of 28 would seriously compromise that.
- Badges
- Science topic