- Isaac and Michael (1995) suggested that “samples with N’s between 10 and 30 have many practical advantages” (p. 101), including simplicity, easy calculation, and the ability to test hypotheses, yet “overlook weak treatment effects.”
- For similar reasons, Hill (1998) suggested 10 to 30 participants for pilots in survey research.
- van Belle (2002) suggested that researchers “use at least 12 observations in constructing a confidence interval” (p. 11).
- In the medical field, Julious (2005) reiterated that “a minimum of 12 subjects per group be considered for pilot studies” (p. 291).
- Treece and Treece (1982), referring to piloting an instrument, noted that for a project with “100 people as the sample, a pilot study participation of 10 subjects should be a reasonable number” (p. 176) but were not clear whether this meant 10 cases or 10% of the project sample size.
- Source Johanson, G. A., & Brooks, G. P. (2009).
https://www.ucl.ac.uk/ich/short-courses-events/about-stats-courses/stats-rm/chapter_1_content/Sample_SizeArticle Considerations in determining sample size for pilot studiesArticle A tutorial on pilot studies: The what, why and HowArticle Initial Scale Development: Sample Size for Pilot StudiesArticle Sample size calculations, pilot studies and confidence inter...Article Determining sample size when assessing mean equivalenceArticle Reviews on the Determining Sample Size using Statistical MethodArticle A simple formula for the calculation of sample size in pilot studiesArticle Pilot studies: Are they appropriately reported?Article A Sample Size Calculator for SMART Pilot StudiesArticle Requirements for Minimum Sample Size for Sensitivity and Spe...
It depends why you are doing the pilot study. So when the big testing companies run pilots they are estimating various parameters for the items (and a bunch of other things) so will have hundreds of thousands of people. Any sources that say there is a single correct range should be judged with skepticism.
Sandeep -
For continuous data and for proportions, one determines the sample size needed in the full project survey by estimating what sample will give you desired standard errors for means, totals, or proportions, given sample design and population or stratum or subpopulation standard deviation(s). So, a pilot study could be used to decide upon how (and if) to subdivide the population, and to estimate population (subpopulation or stratum) standard deviation(s).
Note that a population standard deviation is a constant to be estimated, but a standard error is reduced with larger sample sizes.
Therefore, one might best determine a pilot sample size iteratively, where you estimate standard deviation(s) and keep adding to the pilot sample (randomly according to design, or sampling according to a model) until the estimate(s) of standard deviation(s) stops changing substantially. (The pilot study could also help work out practical problems in data collection, so if that is questionable, then you may need more data.)
Experience and a knowledge of the data for a similar situation (as I believe is noted in Cochran (1977), Sampling Techniques, Wiley) could be used to at least give you a starting point, but if you can start with a smaller pilot sample and add to it until estimated standard deviation(s) seem stable, that might be best. This sample size, as indicated by Daniel, may be vastly different in different applications.
For other kinds of surveys, such as qualitative surveys, with which I am not familiar, I expect that the principle is basically the same: A pilot sample should be big enough to learn how you might want to subdivide your population (or not), and be familiar with (or at least aware of) the variability of the information. Then the full sample should be big enough to feel you can describe that variability adequately.
Cheers - Jim
PS - A pilot study may help to show you when it is best, say in a quantitative survey, to use a more complicated design, say two-stage sampling, cluster sampling, unequal probability sampling, or to use auxiliary/regressor data for model-assisted design-based or model-based methods. This might make more research and perhaps a second pilot study desirable. Anyway, you might want to look at two or more of the following and/or other books:
Thompson, S.K.(2012), Sampling, 3rd ed, John Wiley & Sons.
Cochran, W.G(1977), Sampling Techniques, 3rd ed., John Wiley & Sons.
Blair, E. and Blair, J(2015), Applied Survey Sampling, Sage Publications.
Lohr, S.L(2010), Sampling: Design and Analysis, 2nd ed., Brooks/Cole.
Särndal, CE, Swensson, B. and Wretman, J. (1992), Model Assisted Survey Sampling, Springer-Verlang.
Snijkers, G., Haraldsen, G., Jones, J., and Willimack, D.K.(2013), Designing and Conducting Business Surveys, John Wiley & Sons, Inc.
Brewer, KRW (2002), Combined survey sampling inference: Weighing Basu's elephants, Arnold: London and Oxford University Press.
An Introduction to Model-Based Survey Sampling with Applications, 2012,
Ray Chambers and Robert Clark, Oxford Statistical Science Series
Finite Population Sampling and Inference: A Prediction Approach, 2000,
Richard Valliant, Alan H. Dorfman, Richard M. Royall,
Wiley Series in Probability and Statistics.
Survey Sampling: Theory and Methods, First Edition, 1992,
Chaudhuri, A., Stenger, H.,
Marcel Dekker, lnc., New York, Basel, Hong Kong.
or
Survey Sampling: Theory and Methods, Second Edition
Arijit Chaudhuri, Horst Stenger
March 29, 2005 by CRC Press
Hansen, M.H., Hurwitz W.N. and Madow, W.G. (1953). Sample Survey Methods and Theory (2 vols.) (Republished 1993) Wiley, New York.
Kish, Leslie, 1965, Survey Sampling, John Wiley & Sons, Wiley Classics Library Edition Published 1995
Raj, D., Sampling Theory, 1968, McGraw-Hill
Sukhatme, P.V. (1954), Sampling Theory of Surveys with Applications, The Iowa State College Press
..........
Also, the following may be of interest:
Ken Brewer's Waksberg Award article:
Brewer, K.R.W. (2014), “Three controversies in the history of survey sampling,” Survey Methodology,
(December 2013/January 2014), Vol 39, No 2, pp. 249-262. Statistics Canada, Catalogue No. 12-001-X.
http://www.statcan.gc.ca/pub/12-001-x/2013002/article/11883-eng.htm
And for one stratum or a single subpopulation or unsubdivided population, the following is for the classical ratio estimator model:
https://www.researchgate.net/publication/261947825_Projected_Variance_for_the_Model-based_Classical_Ratio_Estimator_Estimating_Sample_Size_Requirements
SAMPLE SIZE DEPENDS ON THE POPULATION UNDER STUDY. THE MORE THE SAMPLE SIZE APPROACHES THE POPULATION THE BETTER FOR THE STUDY, AND THE MORE COMFORTABLE THE RESEARCHER IS FOR GENERALIZATION OF HIS FINDINGS. THE SAMPLE SIZE CAN BE DETERMINE STATISTICALLY USING "YARO YAMANE" FORMULA FOR FINITE POPULATION. THE FORMULA IS AS FOLLOWS: n = N
__________
1+N(e)2
WHERE; n = SAMPLE SIZE
N = FINITE POPULATION
e = LEVEL OF SIGNIFICANCE(OR LIMIT OF TOLERABLE ERROR)
1 = UNITY (A CONSTANT)
This subject is included in Pilot Plant Scale up technique . You can collect an idea
of any standard book for Project Engineering / Industrial Pharmacy .
I think if you employ quantitative study, n=30 sample size is better to test pre-survey
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size