Frank -

There is no universally applicable number. To even guess what might be acceptable there is much more information that would need to be known: type of data, methodology (which may lead to many more questions), preliminary standard deviation for quantitative data, and goals. If your sample size is too small for your needs, or your plan is inadequate, then even 100% response will not do what you need to do.

Note, for example, that if you have a strictly design-based (i.e., probability of selection) quantitative sampling plan, and there is any nonresponse, then you are violating your plan. How much is too much? You could do a sensitivity study to simulate some cases, but the answer is going to be vague. However if you have a model-based approach to imputation, you may be fine, even with a good deal of imputation, as long as the sample size is not too small, as measured by variance, possibly using the estimated variance of the prediction error for the predicted total.

There are many other possible scenarios. The circumstances for your question need to be explained, preliminary information obtained, and then some rough guesses from ResearchGate respondents might be possible.

For qualitative data, I imagine the answer is going to be very vague/uncertain.

It isn't a straightforward question and answer situation.

Best wishes - Jim

Hi Frank,

Interesting discussion here. Survey research are usually cross-sectional in nature. However, I agree with James that we need to know further description of your methods. If the design of your study is principally cross-sectional in nature, the required sample size could be calculated in two ways, based on your population, if it's a finite or infinite population. If you are using just 1 or 2 industry, your population is static (the workers do not change over time), hence you may use a statistic calculator to define your expected "finite" population. On the other hand, if your target population is entirely big, and you can't ensure the number of people in your population (infinite), you may calculate your sample size based on previous prevalence rate from literature. There are numerous calculators and statistical software that could be helpful to calculate for you. STATA may be able to do.

There is no requirement, but at least half is easiest to justify.

Hope this helps

Matt

Hi! It depends on the requirements set by the analysis you want to perform. You may e.g. have a hard time justifying clusters with less than 30 objects. Is your research descriptive or do you work inductively with shaping models etc.

That is pretty much independent of the industry!

BR, TS