the main problem of course ... with online surveys ... or, with any other
'kind of convenient' surveys ... is to whom do you generalize the results?
the notion is simple in inference ... our sample is meant to tell us
something about THE population that we want to generalize the findings TO
but, to whom is the population to which a survey filled out online ... is
generalizable? we don't know ... though, it MIGHT be safe to say that those
who filled it out online ... probably had a little more techie savvy than
generally speaking
now, in some cases .... this matters not. for, what if i do a survey online
... where the only real interest is to find out IF some people ... use
eudora as their email client software package ... not knowing if
anyone does (other than me) ... so, i ask this question via an internet poll
and, i find that out of about 124 responses i get back ... some 19 said
they did ... and (since the survey probes a bit more) i know this better
since i also ask them a few ??? that they would only know IF they had been
using or had access to eudora ...
the answer to my question then is: yes, some people do use eudora ... and i
found that out by this convenience sample ... and online to boot
of course, if no one says yes ... i still am in a quandry ... for the lack
of a positive response does not prove that 0 use it ...
BUT, if i wanted to use the proportion of 19/124 ... as an estimate of the
population (which we can't really define very well) proportion ... we are
in BIG trouble ... THAT is where the lack of control over the sampling will
reach up and bite you on your mouse!
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================