Ds. Bennett et Ac. Stam, Research design and estimator choices in the analysis of interstate dyads - When decisions matter, J CONFL RES, 44(5), 2000, pp. 653-685
Researchers face three basic questions when testing theoretically driven hy
potheses. First is research design: for example, what population should be
analyzed, what sample should be drawn from that population. and what cases
should be excluded from the sample? What statistical estimator should one u
se? What set of control variables should one employ? Results drawn from sev
eral permutations of the Correlates of War Militarized Interstate Dispute d
ata set are compared using three statistical models. For some theories and
variables (international institutionalism-intergovernmental organization an
d alliance membership; realism-balance of power; expected utility theory-in
ternational interaction game equilibria), research design and estimator cho
ices substantially influence the findings. For others (trade and democratic
peace theory), the relationship between the dependent and independent vari
ables is relatively impervious to research design and estimator choices.