thanks again!

No, sampling “error” is never a concern with qualitative research because no probability sampling techniques are used.

You do not generalize the same way with qualitative research. Instead of measuring the difference between your sample’s results and the population’s, that is, instead of calculating a confidence interval, you use the qualitative notion of “trustworthiness.”

Trustworthiness was developed to complement the qualitative approach by Denzin and Lincoln (editors of the Handbook of Qualitative Research). It suggests that “generalizability” is a function not of math but of “confirmability” (can this be confirmed by other researchers), “credibility” (do your participants “buy” what you’ve said?), “transferability” (are there dominant themes that could be appropriate for other contexts?) and “dependability” (would other researchers likely find the same themes that you did?).

As you can see, the only way to find trustworthiness is to go back to your participants and/or go to other researchers. It’s time consuming (not as easy as just calculating a confidence interval, huh!).

But how do you know when your sample size is big enough? You find what they call “saturation,” which is when you start to get the same kind of answers. This usually happens around the 5th or 6th interview.

So there you have it: more time, energy, effort, and MONEY! Man, qualitative research is such a rip-off! Ah…so if you want to generalize to the population? Do a quant survey AFTER you do your interviews. You can do some confidence calculations then.

1. even with the researcher playing the important role of interpreter in qualitative research, the paradigm is still that the observations and associated insights can be generalized, right? with that in mind, it seems like sampling error is still a concern. for example, if i want to learn about how to design for truck drivers, and i do a ridealong with a single truck driver (a typical technique for product design research), i am assuming that what i observe is representative of truck drivers in general. needless to say, though, this is not necessarily true. therefore, it seems to me that sampling error is an issue, and that the way to avoid is it through a larger sample size.

2. if my first point is correct, and sampling error IS a concern, then what is an acceptable sample size? we can’t do level of confidence calculations, so what standards can we use to answer the question?