26
9
Chiozza and Goemans (2004) divide authoritarian governments into mixed and autocratic,
based on thresholds within the Polity scale.
10
Because of the interaction, the non-interacted conflict variables are not modified when party is
zero and democracy is zero. When this is the case, the effect is represented for military regimes.
11
Essentially, these simulations use the reported coefficients and standard errors to estimate the
effects of the independent variables while holding others constant. In this case, we alter the
values for the interaction variable. See Brambor, Clark, and Golder (forthcoming) for more
details. Specifically, the simulations show that the confidence intervals (2 standard deviations
around the mean) always included zero.
12
This result holds even after we examine the same type of simulations we did for the party
interactions.
13
Because it is not theoretically interesting, we do not report the value of the frailty term.
Convention