This discussion forum is an incredible resource! Instead of independently
responding to everyone who has contributed to the conversation thus far,
I'll issue my response here:
Comments on too-large networks:
Bill's comment on network density refines my motivation for specifying
network parameters required for inclusion in my study because it is not
necessarily network size that I'm concerned about, rather it is the density
of the network. Asking respondents to scroll through a long list of names
should not cause fatigue, but I am concerned about respondents filling out
a set of four multi-item measures for each person selected. But, because I
see no way of predicting network density before I collect my data, I'll
have to 'take the plunge' and hope for the best.
So far there is a consensus, with the exception of network density
concerns, that networks > 120 shouldn't cause me to fret about respondent
fatigue-- as long as I make my measure easy to complete. This is great
news because my committee will be less inclined to challenge me on deciding
to include the larger networks.
Comments on too-small networks:
The point of whole versus ego networks is a very useful distinction to
make. My study entails whole network data on networks that have a mostly
clear organizational boundary.
I respect the comments that a network of 10 is not too small to address
certain questions, but I hope to find networks where not all members share
the same 'horizon of observability' [Friedkin] (thank you, Valdis, for
tossing in this term). My study requires variation in observability, and I
hope to capture that with a wide range in centrality scores.
In light of the responses to my question, then, I will set a minimum
network size to 20 and I will not set a maximum size.
If my study (including the pilot) produces anything that can contribute to
this issue, I'll gladly report it here. It is the least I can do to