***** To join INSNA, visit http://www.sfu.ca/~insna/ *****
Let's see if I understand your situation:
You have some seriously incomplete data.
You think a small fraction of the links are reciprocal.
You don't know about the rest and you don't have enough of the data to
do more than guess.
The people you work with say that the direction of the links must be
included in the analysis.
You want to know whether you should:
-- use the data you have in spite of its incompleteness, but use it as
it is with no distortions, treating it as if it accurately describes the
reality of the situation, even though you know that it actually doesn't;
-- use the data you have, filling in the missing parts on the basis of
"personal assessment", interpreting the results as if they were based on
-- acknowledge that the data's incomplete nature means that you can't
make an empirical determination of directionality of links, and conclude
that it doesn't make sense to use directionality because you simply
don't have the data you'd need to do so.
You want to know whether lack of directionality will skew the metrics.
It will surely skew some; it might not skew others. You could take an
approach similar to the SRDS (Systematic Rejection of Deviant Subjects)
approach a professor I once knew was thought to use to "clean up" his
data. (It's amazing how clean you can make your data appear to be and
how strong your correlations become with the judicious use of SRDS.) In
this case you wouldn't be rejecting "deviant subjects" -- you'd be
rejecting the deviant directionality or lack thereof, thereby making the
data fit the requirements of the party line. Once you've taken care of
that pesky incompleteness, it's smooth sailing, but you'll have some
skeletons in the closet. That's only bad if someone discovers them
....... or if you open the closet door too quickly, having forgotten the
skeletons, and they fall out and knock you on the head, causing some
nasty bruises. What will you say when someone asks you what happened?
Personally, I'd rather take the third approach. I think honesty is
generally the best approach when you're trying to answer a question on
the basis of analysis of some data.
alex chopan wrote:
> ***** To join INSNA, visit http://www.sfu.ca/~insna/ *****
> Fellow SOCNETizens,
> I would like your thoughts on some particular questions RE
> directionality in
> a support/oppose network based on 5 primary nodes...
> I am working with a network that is based on extremely limited
> information -
> and there is no changing that...The network has 5 primary nodes, w/ each
> node having links to a/b 50-100 sub-nodes.
> Due to the paucity of information that I am working with, , I have only
> enough direct evidence/information to establish a link from A to B in a/b
> 80% of the relationships (and a majority of these are 'low certainty'
> 'high certainty'), but no direct evidence to establish a reciprocal link
> from B to A. Based on the information in a/b 25% of those cases, I could
> concievably assess the relationship is mutual (and that is an
> So basically, 65% of my information only informs me of a unidirectional
> So, I see myself as having 3 options:
> 1) Base directionality on evidence - in which case I would only have
> corresponding links between two nodes for under 25% of the nodes.
> This, I
> feel, would only tell me the limits of the information I am working with,
> the gaps of the information on the network, etc...and this potential
> is one that is already known in general, and is not the question I
> have been
> asked to answer...
> 2) Base directionality on personal assessment - in which case the
> of my analysis is diminished b/c the assessments will be subjective,
> and may
> not hold up to scrutiny of reviewers, and therefore not be treated as a
> useful study...
> 3) Forego the use of directionality all together, unless I have direct
> evidence of a non-mutual relationship (A supports B, but B opposes A).
> The people I work with stick to the party line that directionality is a
> must, that lack of directionality in a network analysis will skew the
> metrics, etc...do all/some of you agree? Which option should I take -
> 1, 2,
> or 3?
> Your thoughts would be greatly appreciated.
> Alex Chopan
> Stop worrying about overloading your inbox - get MSN Hotmail Extra
> SOCNET is a service of INSNA, the professional association for social
> network researchers (http://www.sfu.ca/~insna/). To unsubscribe, send
> an email message to [log in to unmask] containing the line
> UNSUBSCRIBE SOCNET in the body of the message.
SOCNET is a service of INSNA, the professional association for social
network researchers (http://www.sfu.ca/~insna/). To unsubscribe, send
an email message to [log in to unmask] containing the line
UNSUBSCRIBE SOCNET in the body of the message.