I am working on a project related to online abuse. The structured questionnaire consists of about 50 questions and we would like to pretest the survey questionnaires before actual data collection. What would be the appropriate way to determine the sample number for pretesting?
There is no codified way of determining sample size for pre-testing instruments. The selection remains subjective. However, you can stick to a simple rule of at least 30 participants and above. You can check out this journal article "Sample size for pre-tests of questionnaires" by Delphine Courvoisier and others.
There is no common way of determining it. It depends on researchers' judgment. I think for the pilot study, you can take 15 samples which represents 30% of your population since your sample is 50.
In health surveys, if you intend to complete detailed psychometric testing (such as internal consistently, content, face, and convergent validity) of your newly developed instrument. Then, we generally follow the rule of 8-10 subjects per item of a scale or questionnaire. Practices could differ across disciplines.
Thank you very much Brian Mumba Ahtisham Younas Zakarie Abdi Bade for your reply.
Zakarie Abdi Bade The questionnaire consists of 50 questions and the sample size is about 5500.
I thought the target population was 50. Since the sample 5500 you can take 7-10% as a pretest. Good luck
I thought the target population was 50. Since the sample 5500 you can take 7-10% as a pretest. Good luck
I thought the target population was 50. Since the sample 5500 you can take 7-10% as a pretest. Good luck.
I thought the target population was 50. Since the sample 5500 you can take 7-10% as a pretest. Good luck.
I thought the target population was 50. Since the sample 5500 you can take 7-10% as a pretest. Good luck. Thank you
I thought the target population was 50. Since the sample 5500 you can take 7-10% as a pretest. Good luck. Thank you
The availability of the sample should also be taken into account in the pretest. When evaluating the pretest, it should also always be taken into account that, depending on the sample, certain selection effects and limitations exist that are not actually present in the intended population.
The sample size should therefore also be based on the availability and the time frame of the project.
If you use probability sample size, you may use satistical formula. On the other hand, you use google doc or direct hard copy. Your given information is not clear. How may population size? and What will be statisticla techniques and tools used? Again, it depends on questions types like quantitative or qualitative.
Qualities of a good sample are: large enough, reflective of the target population, etc . By the same token, a pretest is like a micro-sample. In that case, the number need not be too high. 10/25 may be considered as the lower limit for small to large population. Any thing above 50 for a pretest is likely to be unnecessary.
Thank you Yekeen Abdul-Maliq and Sushanta Roy Chowdhary for your suggestion.
what is your total population size? so that I can set target respondent amount and the number of respondent required to undertake your pretest.
It depends on what you want to test. If you want to test data collection methods, such as the impact of offering an incentive or using a special contact procedure, then you need quite a few. If you want to test the survey content, then I find the value of additional pretests quickly diminishes. For surveys sponsored by the U.S. Federal government, people often test with nine (because testing more would require OMB approval), and that tends to work well. I've found that people often repeat each other after about 5 or 6, particularly if you try for diversity in who you sample.
Thank you very much Bradford Chaney for your reply. We want to test both data collection methods and the content of the survey questionnaires.
Since there’s no single way to do it, it depends on your objective of protesting. Do you want to test the tools or do you want test whether the sample represent the population?. If you just want to test whether the tools are applicable to your research question, you can take 6-8/50 to test. But, if you want to test whether the sample is generalizable, I think you may need a larger pretest about 30% of your sample.
- Badges
- Science topic