IMO, after having taught applied multivariate stats for over 10 years Hair et al is worthless. The standards for this answer are: https://www.bing.com/search?q=Cochran+Sampling+Techniques&cvid=0772bd822dda41649b5aa9acece3d702&FORM=ANAB01&PC=U531

and

https://b-ok.cc/book/2324520/acf6e2

You could also use: https://en.wikipedia.org/wiki/G*Power

Best, D. Booth

It may be seen in the following book

Business Statistics by examples.Fifth edition by Terry Sinich

IMO, after having taught applied multivariate stats for over 10 years Hair et al is worthless. The standards for this answer are: https://www.bing.com/search?q=Cochran+Sampling+Techniques&cvid=0772bd822dda41649b5aa9acece3d702&FORM=ANAB01&PC=U531

and

https://b-ok.cc/book/2324520/acf6e2

You could also use: https://en.wikipedia.org/wiki/G*Power

Best, D. Booth

Mulusew Andualem Asemahagn

https://www.researchgate.net/post/Is_there_a_rule_of_thumb_formula_for_sample_size_calculation_comparing_2_rates

https://www.quora.com/What-is-a-good-rule-of-thumb-for-determining-sample-size-for-a-survey

https://www.jstor.org/stable/1267178

Determination of sample size will be informed by the research design whether it is qualitative or quantitative. You could refer to this book

Research design: Qualitative, quantitative, and mixed methods approaches

by JW Creswell, JD Creswell - 2017 - Sage publications

You all, thank you much for your valuable responses.

If I get your question right, a rule of thumb for choosing the sample size is just your guess of how many samples to take for estimating the value of a population parameter. If you do not have information on the population variability you could chose a sample size as a "rule of thumb" based on previous knowledge of similar populations or of your own knowledge of the population to study. Popular rule of thumb is the sample size n = 30 observations which is recomended in many studies and even books in the Life Science discipline. There are several accounts of where this "magical" number comes from, but that is a long debate. Usually, for small populations larger samples intensities must be taken relative to large populations. More than thinking of rules of thumb, if you lack information of population variability, you should consider to take a number of samples based on the difficulty of measuring your variable(s) of interest, and the time and money available for your study. Also how precise your estimation is needed.. For example, if your population is N = 10^6 individuals and you yourself have to go to ask their opinion on question X, then 20-30% of the population does not sound as a good rule of thumb. Maybe, just 0,001% of intensity is good enough to get a reliable estimate.

Best Regards

A rule of thumb refers to the ratio between DV and IV. There are many differents rules of thumb, starting from 5 cases per 1 DV up to a sample size of 300 with not too many DV:s.

Mulusew -

"Rules of thumb" can have no merit unless they are in a defined, perhaps narrow context. I'd avoid them.

For sample size, there are three key components: population standard deviation (a constant for each population, subpopulation, or stratum), desired standard error, and methodologies (sampling scheme and estimation or perhaps prediction from regression). - (Note that below, for regression, the standard deviation, sigma, for estimated residuals, is complicated by heteroscedasticity, but one can factor estimated residuals into random and nonrandom factors. The former can be used for finding a constant sigma part, and the latter is related to regression weight. Note that number of independent variables only impacts degrees of freedom.)

Under a given methodology for sampling and a methodology for estimation or prediction, what sample size causes the necessary standard error(s) considering the existing standard deviation(s)?

For more than one, say, question on a survey, you need to do this for each important case and consider the largest result. To obtain standard deviations, you may need a pilot study if you do not already have enough information in that regard. Also, the more you know about the population, and any possible auxiliary data for a possible model-assisted approach or as regressor data for a possible model-based approach, otherwise a strictly probability-of-selection-based approach, even information to stratify well, the better you can decide on appropriate methodologies.

For modeling, sigma for the estimated residuals may be thought of as the standard deviation in question. There is also variance from the model to consider such that the square root of the estimated variance of the prediction error is what is then needed for standard error. In finite populations we can find the estimated variance of the prediction error for a predicted total. If looking at individual predictions, then results vary by prediction, especially if there is heteroscedasticity, which is a natural phenomenon. Graphical residual analyses can help to see how well you are doing overall with model fit, and a cross-validation can be used to check if you have overfit to a given sample.

So you see that a 'rule of thumb' here is often a gross oversimplification.

Perhaps you should start with a pilot study to learn about your situation, and plan from there.

Nancy mentions qualitative designs, which I would also guess could benefit from a pilot study.

Best wishes - Jim

Dear Syed Azamussan, what is the recommendation of Joe Hair?

Following

https://scholar.google.com/scholar?oe=utf-8&gcc=my&ctzn=Asia/Kuala_Lumpur&ctf=0&v=12.1.9.23.arm64&fheit=0&biw=360&bih=719&ntyp=1&ram_mb=7614&devloc=0&gs_lp=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&cct=4324&client=ms-android-oppo-rev1&wf=pp1&padt=200&padb=720&hl=en-GB&cs=0&cds=2&psm=1&um=1&ie=UTF-8&lr&q=related:FODJDwf-ywoQGM:scholar.google.com/#d=gs_qabs&u=%23p%3DFODJDwf-ywoJ