In chapter six of Salsa
Dancing into the Social Sciences, Kristen Luker explains how one of the
main reasons to perform sampling is to come up with generalizations (e.g. if
six out of ten people react a certain way in a random sample, it means that
sixty percent of the population is likely to do so as well) (Luker 2008). While
these generalizations seem to be substantiated by the sampling process, I am
still troubled that we put so much faith into sampling to make these statements.
What if the random sample is somehow tainted? What precautions are taken to
make sure the sample is truly random?
In this chapter, Luker also explains the concept of data
outcropping (2008) which involves sampling a group of people relevant to the
research issue (e.g. sampling a group of pregnant women for research on
pregnancy vs. sampling an entirely random population where some of the
participants’ answers will not apply to the research). If data outcropping does
help make sampling results more accurate, does this mean that generalizations
that are a result of data outcropping should be trusted more than other
research generalizations? Also, are we expected to always perform data outcropping
or does the type of research we’re doing dictate the way that we would sample?
Luker, K. (2008). Salsa
Dancing into the Social Sciences. Cambridge: MA: Harvard University Press.
I am no stats-wizard (all I have is one social science stats class under my belt) but I am pretty sure they have ways of making sure a sample is truly random. They also have ways of making sure outliers don't skew the results. I am sure another member of the blog can speak more to this...
ReplyDeleteAs for the data outcropping I don't think we can make statistical claims about the population we isolated but we can make logical claims. In your example, Luker would argue that there's no logical reason that the pregnant women you talked to aren't typical pregnant women. However they aren't statistically representative of all pregnant women in America (Luker, p. 125). This has to do with sampling, because the women you talk to won't truly be randomly selected. They might all from a local doctor's office with which you are affiliated. This doesn't make it more or less reliable because they are trying to prove different things.
I think the type of sampling you do will very much depend on your research question and the scale of your research. For our purposes data outcropping works well because we don't have the time or the resources to evaluate mass quantities of data.
I hope this made some sense Amy! :)