Interpersonal relationships are an important and integral part of numerous
social science research agendas. Analytical tools have been created in the
last 10 years that model dyadic interactions. In particular, this article f
ocuses on the dyadic models of Fienberg and Wasserman [Fienberg, S.E., Wass
erman, S., 1981. Categorical data analysis of single sociometric relations.
In: Leinhardt, S. (Ed.), Sociological Methodology. Jossey-Bass, San Franci
sco.], Holland and Leinhardt [Holland, P.W., Leinhardt, S., 1981. An expone
ntial family of probability densities for directed graphs. Journal of the A
merican Statistical Association 76 (1981) 33-51.], Iacobucci and Wasserman
[Iacobucci, D., Wasserman, S., 1988. A general framework for the statistica
l analysis of sequential dyadic interaction data Psychological Bulletin 103
(1988) 379-390.] and Wasserman and Iacobucci [Wasserman, S., Iacobucci, D.
, 1986. Statistical analysis of discrete relational data. British Journal o
f Mathematical and Statistical Psychology 39 (1986) 41-64.]. However, measu
rement issues like reliability and validity, as discussed by Alien and Yen
[Allen, M.J., Yen, W.M., 1979. Introduction to Measurement Theory. Brooks/C
ole, Monterey, CA, 1979.], Nunnally [Nunnally, J., 1978. Psychometric Theor
y, 2nd edn. McGraw-Hill, New York, NY, 1978.] and Uebersax [Uebersax, J.S.,
1988. Validity inferences from interobserver agreement. Psychological Bull
etin 104 (1988) 405-416.], have not been considered in conjunction with the
se models, and little is known about the empirical performance of the dyadi
c models under sub-optimal measurement quality conditions. We offer two ess
ential approaches to ascertaining the level of measurement error in the obs
erved indicators of social ties and relationships. The first approach combi
nes latent class and social network models in one integrated framework: and
allows for the simultaneous study of measurement and dyadic structural iss
ues. The second approach is an alternative that may be more useful to socia
l science researchers, both because the method is more accessible and becau
se researchers could apply the techniques to data they have already partial
ly analyzed. This approach is a two-staged procedure whereby in the first s
tage, a probability model based on latent class analysis is estimated which
provides an indication of the measurement quality in the data In the secon
d stage, traditional social network models are estimated. To investigate th
e implications of different levels of measurement error for interpreting th
e nature of the network ties and the dyadic parametric performance, we also
designed a Monte Carlo experiment. Measurement error is simulated as the l
ikelihood of a binary relational choice (for simplicity) being inaccurately
classified, where incorrect diagnoses can result from poor interitem agree
ment (i.e., unreliability) or poor interrater agreement. The simulation can
be used by researchers in combination with the two-stage approach. The res
ults of the simulation provide guidelines for situations when social networ
k models can withstand a reasonable degree of sub-optimal measurement quali
ty and highlight adverse conditions which can significantly affect the perf
ormance of the modeling approach. Further, the simulation shows that sample
size assists in reducing the chances of making Type II errors, but it does
not compensate for biases in parameter estimates in the presence of increa
sing error. Finally, the measurement and dyadic analytical methods are appl
ied to a real dataset describing interorganizational relational activity us
ing multiple raters. Recommendations are offered to guide the researcher in
making decisions about research design when using dyadic models.
(C) 1999 Elsevier Science B.V. All rights reserved.