for example, total respondents are 100. 50 answered Q6, 45 answered Q7 , 60 answered Q8. What should I do to write the results to be acceptable for the journal reviewers?
The survey has reliability of 95%.
You should not be worried about that. This is the nature of human, 99% of the surveys are like that.
It is fine. The issue could be if your sample would not representative. You can check out this by different methods.
One of the easiest, compare the sample demographics with the population demographics (age, education, income, ethnicity, gender, etc.). Do t-tests. If you cannot find any significant difference, your survey is representative.
If you find, do not worry, use the raking algorithm (iterative proportional fitting) to weight the respondents.
Please find the details of the raking method in my presentation enclosed:
Conference Paper Manage sample and population differences by weighting - usin...
Back to your original question, you can do plenty of things (before raking!):
- segmenting respondents
- grouping respondents and compare the groups by t tests. It is possible that the question skippers are similar or different compared to those who answered all
- for some statistical analysis, you can leave out those who have not responded or you can code their responses differently (e.g. you calculate standard deviation by using n=(n0-n nonresponding), but for other analysis you replace those with the mean).
All of the above should be justified and explained in your paper. Without knowing the case, I cannot advise what would be the best. Do not forget, you can choose any way to analyze your dataset if it makes sense and you explain it.
You can say in your research design that participants were not required to answer all questions. When you report a finding, say something like "of those who responded, 38 (70%) said that......."
This one's not difficult, Ms. Gamee. If the questions are substantially independent (that is, yielding replies on essentially unrelated issues), then you should have no difficulty simply ignoring the respondants who did not answer those questions. If, on the other hand, the questions are sequential or directily related (that is, one answer depends directly on the previous question), you'd be well advised to disregard the entire set from that respondant. It's simply not producing a reliable array of data points for you. Most researchers are disappointed in that reply, but your real aim here is both reliability and validity. Once you have quotients that meet your research standards, don't worry about volume.
Dear May
The incomplete surveys should be discarded /neglected, and the remaining 50 (>30) surveys are sufficient for statistical analysis.
Good Luck.
- Badges
- Science topic
- Similar topics
- Social Science
- Media
- Journalism