HI
Wanted to know what is the best sample size for a non probability pre test Questionaire.
what is recommended for a minimum sample size of 500?
It sounds like you're using a non probability samplimg scheme. If that's the case then all one can say is more is probably better than less. If not.a better description of your sampling plan is needed. In that case consult Schaefer et al Elementary Survey Sampling downloadable from the z-library. I ff needed please post another question. Best wishes David Booth
Hi,
You have different types in Nonprobability sampling methods and there is no particular sample size given.
Hello Inas,
The answer depends on what specific analyses you intend to perform for this pre-test appraisal of an instrument.
If your intention is to calibrate the items via item response theory, then you should aim for at least 200 cases for a one-parameter IRT model, and more (500-1500) for two- or three-parameter models.
If your intention is to factor the measure, using exploratory factor analysis, then you should aim for 10-20 cases per item (with a lower bound of 100). The reason is that the observed correlations are presumed to be population parameters (they won't be, of course); as well, you're looking for a structure to explain/account for the k * (k - 1)/2 unique correlations/covariances among the set of k items.
If you're concerned about basic item analysis and internal consistency reliability estimates for scores, then you should aim for: (a) more cases than items; and (b) a lower bound of 50-100 cases.
If you have something else in mind, then do follow Professor Booth's suggestion, and post further explanation. I'm confident that the Rgate crowd can then offer more focused recommendations.
Good luck with your work.
Hi
Thank you all for the answer, to explain further, I will use convient sampling and the approach will be CFA confirmatory factor analysis.
Could any provide me with useful references I that could support me with the pre test sample size choice?
Thanks alot.
Hello again Inas,
The glib advice is 10-20 cases per parameter to be estimated in the CFA. For example, each variable-factor loading is estimated. You can, of course, be more specific by evaluating some simulations that mirror the proposed factor structure (see this example: file:///C:/Users/Morse/AppData/Local/Temp/samplesizedeterminationinCFA.pdf)
The concern associated with a convenience sample is that it may not be representative of your intended population.
Good luck with your work.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size