*****  To join INSNA, visit  *****

In the context of an public school administrative network, I am trying to
analyze the relationship between actor-level characteristics and 1) egocentric
network structure and 2) centrality.  N=34 adminstrators.

IV:  Characteristics
    PERSONAL (gender, age),
    ADMINISTRATIVE POSITION (administrative level, school  level), and
          --PROXY for power/authority
    ADMINISTRATIVE EXPERIENCE (years in admin, number of schools as
          --PROXY for time & familiarity across network

DV:   Network Structure
     EGOCENTRIC NETWORK STRUCTURE (size, density, % bidirectional ties, mean
strength of ties)
     NETWORK CENTRALITY (betweenness, closeness & eigenvector)
                                             (not degree as this is bascially
being used for size)

Previous analysis of this network has been done at the dyadic level (max
n=1122).  However at the actor level, the n=34.  Assuming a case study approach
where the entire population is included (no random sampling), what is the best
way to demonstrate relationships between variables?

    Creating categories, I have done this using t-test and ANOVA.
    I have also used correlation for the interval data.

However, it seems like it would be much beneficial (and succinct) to present
regression models.  My question is how will a regression analysis with N=34
(non-random / whole population) be critiqued and what is the best way to present
it to respond to these potential critiques (e.g. small n).

Clearly the findings would only be descriptive of this network and not
generalizable to other networks.  Would presenting it as a descriptive analysis
solve the small n?  I would appreciate any suggestions for how to approach and
present this analysis.

Julie Hite
Brigham Young University
[log in to unmask]

SOCNET is a service of INSNA, the professional association for social
network researchers ( To unsubscribe, send
an email message to [log in to unmask] containing the line
UNSUBSCRIBE SOCNET in the body of the message.