On Mon, 08 Apr 2013 07:58:30 -0700, David Epstein wrote:
On Mon, 8 Apr 2013, Rick Froman went:
As to the snowball technique, it is quite common in qualitative
studies where the "representativeness" of the sample is irrelevant
(and it is very difficult to find many participants so you rely on
the few you find recruiting their own friends. Qualitative research
has no concept of a larger population of which this sample is a
representative microcosm. All that matters is that you correctly
communicate the experience of those that end up in your
sample. Thinking that such a sample could serve as well for the
quantitative research, which has entirely different assumptions,
seems to indicate that they are really just using the "it's just
qualitative" fall-back when their quantitative data is questioned.
And as the original poster, I want to emphasize that I greatly value
qualitative research. Focus groups, interviews, and ethnographic
immersion can provide knowledge you would never get from other
approaches. (I'm a coauthor on a focus-group paper, and I'm quite
proud of it, even though no one ever cites it.) You just need to
write clearly about what you're doing--and speak up real loud if it's
misrepresented in a press release or a spate of mass-media coverage.
I have not been following this thread as closely as perhaps I should
have but there are a few points that concern me that have not been
adequately addressed:
(1) Allen Esterson mentioned that in response to his queries, the
author of the research mentioned that he used "snowball sampling".
For those who need a brief review of sampling techniques and their
impact on analysis, I suggest taking a look at Kakinami & Conner (2010)
"Sampling Strategies For Addiction Research", a copy of which can
be obtained here:
http://dl.ravannews.com/files/Addiction_research_methods.pdf#page=44
Chapter 3 is on sampling techniques and the major types of sampling
schemes (i.e., probability, non-probability which snowball sampling is
an example, and qualitative sampling). I think that there is some confusion
in keeping non-probability sampling separate from qualitative sampling
in this discussion and this reference will help to highlight it.
(2) Non-probability sampling like snowball sampling is typically used
with "hard to reach" populations, such as substance abuse users like
crack smokers -- one can't just start calling people on the phone and
ask if they use crack, one typically asks one crack user to identify another
crack user who could participate in the study. Similar situations hold
with sexual activity (e.g., HIV spread, other STDs; network analysis
comes in handy for this type of analysis as well as the recognition that
people in such a sample are "correlated" or clustered because of
the dependency of a selected person on the person who nominated
him/her). Snowball sampling leads to problems in analysis which specialized
software (e.g., SUDAAN) or procedures (e.g., multilevel modeling)
can attempt to deal with but, if the researcher has a choice "Respondent
Driven Sampling" (RDS) a variation of snowball sampling where the
participant is asked to give a certain number of potential participants
(see Kakinami & Conner above). More statistical analysis is possible
here. All this being said, it is unclear to me that the chi-square analyses
shown earlier in this thread are valid since one does not know how
the sample represents the population it was drawn from.
(3) It could just be me (based on my experience with drug and HIV
research) but I don't think that participants for the reported study
represent
a "hard to reach" population. Indeed, census or similar types of data
sources should allow one to locate families with or without children
who can then be sampled and interviewed or given questionnaires to
fill out. The rationale for using snowball sampling is not at all clear to
me and may actually represent a confusion about what type of sampling
was actually used.
So, if anyone can clarify, I'd appreciate it.
-Mike Palij
New York University
m...@nyu.edu
---
You are currently subscribed to tips as: arch...@jab.org.
To unsubscribe click here:
http://fsulist.frostburg.edu/u?id=13090.68da6e6e5325aa33287ff385b70df5d5&n=T&l=tips&o=24854
or send a blank email to
leave-24854-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu
