Elizabeth -

I think that what you have in mind is the difference between inferring from a sample to a population, as opposed to descriptive statistics of data collected, from which you are not inferring to a population.  This might occur for a census, that is, where you already have the population.  If you just want to describe the data set, then you do not worry about inference.

For inference, commonly one uses design-based sampling and estimation methods, which means you need to know probabilities of selection (from the population of interest).  Such designs include stratification, cluster sampling, and multi-stage versions.  To do better than simple random sampling, one must know something about the population.  If you have detailed auxiliary data, you can use model-assisted design-based sampling and estimation.  Prediction-based methods (i.e., based on regression) may also be of use. 

Here are some textbooks which might be of interest to you:

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.    

As an introduction to the different ideas available for inference, consider the following historically instructive and entertaining 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

Cheers - Jim

Just a quick addition to Jim's comments. I would always use a probability sample unless I knew for sure that I will not be using statistical inference. Better to be prepared than to have to take an extra sample. Best, D. Booth

I agree with D. Booth, if you are going to make an inference, probability sampling technique should be utilized in the selection process. Or even using any inferential statistical tools, as much as possible it is a necessary requirement to use probability sampling techniques to select sample participants.

best regard, Elizabeth

Dear Elizabeth Reese

The main consideration is the 'How generalizable are your findings to the target population?', and probability sampling is recommended where this is the goal.

I have encountered ideas that 30 as a sample size is acceptable. In a research on metaphors used in ads, a sample of 30 ads will be analyzed in a descriptive research. May I know your opinion on this?

@carissa, n= 30 was an old incorrect application of the central limit theorem

Sample size should be calculated by something like G*power on line calculator. Best wishes, David Booth