[Air-l] Question re Size of Data Set...

[Matthew Pearson]

Hello All:

This is my debut post to this excellent list after a long time spent  
reading lots of good stuff from others.

I have a question re the size of my data set for my dissertation  
project: Is my data set too large, too small, or just right?

I'd very much appreciate any insight/ideas/feedback anyone has about  
this.  My advisor and I aren't that sure about this issue, and I  
haven't been able to discern much about this issue from a lot of the  
studies I've read.

I do realize my question is thus far meaningless without knowing  
anything about my project, so here's some more information/background:

I'm doing a close look at one message board community--one devoted to  
discussion of a particular college basketball team.  I've got all  
sorts of things I'm interested in, but my central research question  
has to do with the ways that people teach each other and learn from  
one another the conventions for discourse in/on the message board.   
(I'm also interested in potential emerging genres of writing, the  
influence of sports fandom on online literacy practices, and perhaps  
even examining issues related to gender (which I realize is a pretty  
general thing to say, but I'll keep it at that for now).)

I've got two main sources of data: both (1) archived threads/posts  
from the message board, and (2) online questionnaires that  
participants/members filled out.  My question concerns source (1)-- 
the archival data.

I have tons of data archived.  I used one of those "site-sucker"  
programs to grab all the discussions on the message board over about  
a 8 month period of time.  Given that this message board is a pretty  
busy one and that I'm using a ground-theory approach to the data  
analysis, I chose to sample a smaller set of the overall data.  I  
used an "event sampling" method and, with input from posters on the  
message board, chose 5 "big" events around which to sample  
discussion.  I then also chose 5 other events that occurred during  
the months I archived discussion that were not listed as "big" events  
by anyone who offered their sense of the "big" events.  I didn't,  
though, choose just those threads of discussion related to those  
"big" and non-big events, but rather used those as anchoring moments  
in time, and then sampled ALL discussion that occurred on those  
dates, and one day prior and one day later.  This resulted in such a  
large data set that I ended up using only 3 "big" events and 3 non- 
big ones, and then sampling for those dates, and the days immediately  
around them.

What I'm left with now is about 4000 individual .html pages, some of  
which have fairly detailed threads of discussion, with sizable  
individual posts, and also, of course, many of which that have  
cursory, short sentences that perhaps look more like "chat."  This is  
a lot of stuff to wade through, yet it does represent only 18 days of  
life on this message board.  Thus far I've been going through the  
data in separate "passes," looking for answers to particular aspects  
of my research question, and it's a daunting thing.  I know research  
takes a lot of work and time, but I thought it wise to get feedback  
to see if I'm going overboard here.

So does my sample sound reasonable?  I'm well aware that the way I  
sample will directly impact the kinds of conclusions  can draw and  
level of rigor folks see in my work.

Any thoughts?  Good sources re this kind of methodology?  I've got  
Virtual Methods Ed. by Hine, among other sources, and haven't seen  
anything yet re sample size.  Maybe I missed it somehow?

many thanks,

Matthew Pearson
mdpearson at wisc.edu
PhD Candidate, University of Wisconsin Department of English-- 
Composition and Rhetoric;
Research Assistant, UC-Irvine Writing Project;
& Man on the Street