[Air-L] difference in representativeness: offline or online

[Kevin Guidry]

On Fri, Oct 1, 2010 at 10:25 AM, Elke Greifeneder
<elke.greifeneder at googlemail.com> wrote:
>
>
> I am working on how to collect purposeful data in digital library
> evaluations.

(Side note: For many of my colleagues, this would probably be program
evaluation or assessment and not research.  It's usually not a
terribly important distinction but some people insist on making it.)

> My question is whether this produces representative data?

I do a lot of quantitative work, much of driven by survey data, so my
response is framed by those sensibilities.  In much of my work, the
questions of representativeness and generalizability are closely
linked to the phenomenon I am studying.  Representativeness denotes a
context i.e. In *what way* is the sample representative/not
representative of the population?  Since you could answer that
question in an infinite number of ways (gender, age, location, SES,
education level, etc.), you should limit your inquiry to those
characteristics which you know or suspect to be important to the
specific phenomenon you are studying.  So you have to figure out the
personal characteristics that are important to your study and if your
sample differs from your population in significant ways.  And you know
your phenomenon and population better than anyone else here so you're
in the best position to know the key characteristics for your study.

To be completely honest, the fact that you're asking the question
indicates to me that you have at least a suspicion that your sample
may not be completely representative of your population.  Often there
is nothing you can to do to directly address that problem but
sometimes you can even then you can note it as a potential limitation
and use some caution in generalizing your findings.


Kevin