autocracy promotion, conflict, ideology, revisionist states
Part 1: Dyadic Models
Dyadic Models Methods
First set Additional Controls
- Logistic regression: A standard GLM model with additional controls for H1 and H2
- Linear Probability Model: A standard linear probability model with additional controls and interaction effects for H1 and H2
Second Set
- Interaction Models for several possible confounding variables with full controls
Final Set
- Rare Events Logit based on King and Zeng, 2001
- Panel LPM: This is a panel linear model (linear probability model) with twoway, within effects. All additional covariates had to be omitted for the model to not be computationally singular.
- WLS: Panel weighted least squares model with twoway, within effects. Some controls are omitted as unnecessary or harmful in a fixed effects model. The model is weighted by probability of conflict providing an alternative to rare events logit.
- Fixed effects model: This is a panel logistic regression model using dyadic fixed effects.
Logits and LPMs - all models
Show the code
# logit modelh1full <-glm(sidea_revisionist ~ sidea_revisionist_domestic + sidea_national_military_capabilities + sideb_national_military_capabilities + sidea_e_miinteco + cold_war + sidea_revisionist_domestic + sidea_national_military_capabilities + sideb_national_military_capabilities + sidea_military_support + sidea_winning_coalition_size + cold_war + t + t2 + t3 + sidea_muslim, family =binomial(link ="logit"), data = autocracies)h2full <-glm(sidea_targets_democracy ~ sidea_revisionist_domestic + sidea_national_military_capabilities + sideb_national_military_capabilities + sidea_e_miinteco + cold_war + sidea_revisionist_domestic + sidea_national_military_capabilities + sideb_national_military_capabilities + sidea_military_support + sidea_winning_coalition_size + cold_war + t + t2 + t3 + sidea_muslim, family =binomial(link ="logit"), data = autocracies) alternate1full <-glm(sidea_revisionist ~ sidea_dynamic_leader + sidea_revisionist_domestic + sidea_national_military_capabilities + sideb_national_military_capabilities + sidea_e_miinteco + cold_war + sidea_revisionist_domestic + sidea_national_military_capabilities + sideb_national_military_capabilities + sidea_military_support + sidea_winning_coalition_size + cold_war + t + t2 + t3 + sidea_muslim, family =binomial(link ="logit"), data = autocracies) alternate1int <-lm(sidea_revisionist ~ sidea_dynamic_leader * sidea_revisionist_domestic + sidea_national_military_capabilities + sideb_national_military_capabilities + sidea_military_support + sidea_winning_coalition_size + cold_war + t + t2 + t3 + sidea_muslim, data = autocracies) alternate2full <-glm(sidea_targets_democracy ~ sidea_dynamic_leader + sidea_revisionist_domestic + sidea_national_military_capabilities + sideb_national_military_capabilities + sidea_military_support + sidea_winning_coalition_size + cold_war + t + t2 + t3 + sidea_muslim, family =binomial(link ="logit"), data = autocracies) alternate2int <-lm(sidea_targets_democracy ~ sidea_dynamic_leader * sidea_revisionist_domestic + sidea_national_military_capabilities + sideb_national_military_capabilities + sidea_military_support + sidea_winning_coalition_size + cold_war + t + t2 + t3 + sidea_muslim, data = autocracies) stargazer(h1full,h2full,alternate1full,alternate2full, type ="text")
plot_model(h1full, type ="pred", terms ="sidea_revisionist_domestic", title ="GLM model Predicted Probabilities")
Show the code
## plot predicted probabilities for h1alternatefull using sjplot plot_modelplot_model(alternate1full, type ="pred", terms =c("sidea_dynamic_leader","sidea_revisionist_domestic"), title ="GLM model Predicted Probabilities")
Show the code
## plot predicted probabitilies for h2alternatefull using sjplot plot_modelplot_model(alternate1int, type ="pred", terms =c("sidea_dynamic_leader","sidea_revisionist_domestic"), title ="Linear model with interaction effects \n Predicted Probabilities")
Show the code
## plot predicted probabilities for alternate2full using sjplot plot_modelplot_model(alternate2full, type ="pred", terms =c("sidea_dynamic_leader","sidea_revisionist_domestic"), title ="GLM model Predicted Probabilities")
Show the code
## plot predicted probabitilies for alternate2int using sjplot plot_modelplot_model(alternate2int, type ="pred", terms =c("sidea_dynamic_leader","sidea_revisionist_domestic"), title ="Linear model with interaction effects \n Predicted Probabilities")
Show the code
plot_model(alternate2int, type ="pred", terms =c("sidea_revisionist_domestic", "sidea_dynamic_leader"), title ="Linear model with interaction effects \n Predicted Probabilities")
Interaction models with other control / explanatory variables
Show the code
W_int <-lm(sidea_targets_democracy ~ sidea_revisionist_domestic * sidea_winning_coalition_size + sidea_dynamic_leader + sidea_national_military_capabilities + sideb_national_military_capabilities + sidea_military_support + cold_war + t + t2 + t3 + sidea_muslim, family =binomial(link ="logit"), data = autocracies) ## plot predicted probabitilies for W_int using sjplot plot_modelplot_model(W_int, type ="pred", terms =c("sidea_winning_coalition_size","sidea_revisionist_domestic"), title ="Linear model with interaction effects \n Minimal Winning Coalition \n Predicted Probabilities")
Show the code
mil_supp_int <-lm(sidea_targets_democracy ~ sidea_revisionist_domestic *sidea_military_support + sidea_dynamic_leader + sidea_national_military_capabilities + sideb_national_military_capabilities + sidea_winning_coalition_size + cold_war + t + t2 + t3 + sidea_muslim, family =binomial(link ="logit"), data = autocracies) ## plot predicted probabitilies for mil_supp_int using sjplot plot_modelplot_model(mil_supp_int, type ="pred", terms =c("sidea_military_support","sidea_revisionist_domestic"), title ="Linear model with interaction effects \n Minimal Military Support Interaction \n Predicted Probabilities")
Show the code
plot_model(mil_supp_int, type ="pred", terms =c("sidea_revisionist_domestic","sidea_military_support"), title ="Linear model with interaction effects \n Military Support Interaction \n Predicted Probabilities")
Show the code
cold_war_int <-lm(sidea_targets_democracy ~ sidea_revisionist_domestic * cold_war + sidea_winning_coalition_size + sidea_dynamic_leader + sidea_national_military_capabilities + sideb_national_military_capabilities + sidea_military_support + t + t2 + t3 + sidea_muslim, family =binomial(link ="logit"), data = autocracies) stargazer(W_int,mil_supp_int,cold_war_int, type ="text")
# panel linear modelh2plm <-plm(sidea_targets_democracy ~ sidea_revisionist_domestic + sidea_dynamic_leader, effect ="twoways", model ="within", index =c("dyad","year"), data = autocracies)stargazer(h2plm, type ="text")
# fixed effects model h2fe <-bife(sidea_targets_democracy ~ sidea_revisionist_domestic + sidea_dynamic_leader + sidea_national_military_capabilities + sideb_national_military_capabilities + sidea_military_support + sidea_winning_coalition_size + cold_war + sidea_muslim + t + t2 + t3 | dyad, model ="logit", data = autocracies) # h2pglm <- pglm(sidea_targets_democracy ~ sidea_revisionist_domestic + sidea_national_military_capabilities + sideb_national_military_capabilities + sidea_e_miinteco, data = autocracies, effect = "time", model = "within", family = binomial(link = "logit"))summary(h2fe)
binomial - logit link
sidea_targets_democracy ~ sidea_revisionist_domestic + sidea_dynamic_leader +
sidea_national_military_capabilities + sideb_national_military_capabilities +
sidea_military_support + sidea_winning_coalition_size + cold_war +
sidea_muslim + t + t2 + t3 | dyad
Estimates:
Estimate
sidea_revisionist_domestic 1.7121997
sidea_dynamic_leader -0.5672233
sidea_national_military_capabilities 66.4562633
sideb_national_military_capabilities -15.2659627
sidea_military_support 4.9792400
sidea_winning_coalition_size 6.7734465
cold_war -0.4922531
sidea_muslim 79422960859867.6406250
t -0.7307293
t2 0.0517126
t3 -0.0010707
Std. error z value
sidea_revisionist_domestic 0.3320888 5.156
sidea_dynamic_leader 0.1370634 -4.138
sidea_national_military_capabilities 11.5202061 5.769
sideb_national_military_capabilities 4.1459887 -3.682
sidea_military_support 0.8740497 5.697
sidea_winning_coalition_size 1.4576541 4.647
cold_war 0.1815478 -2.711
sidea_muslim 371818342720829.6250000 0.214
t 0.1648707 -4.432
t2 0.0176531 2.929
t3 0.0004707 -2.275
Pr(> |z|)
sidea_revisionist_domestic 0.00000025248 ***
sidea_dynamic_leader 0.00003497364 ***
sidea_national_military_capabilities 0.00000000799 ***
sideb_national_military_capabilities 0.000231 ***
sidea_military_support 0.00000001221 ***
sidea_winning_coalition_size 0.00000337102 ***
cold_war 0.006699 **
sidea_muslim 0.830854
t 0.00000933048 ***
t2 0.003396 **
t3 0.022922 *
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
residual deviance= 1655.19,
null deviance= 2201.78,
n= 3279, N= 87
( 335021 observation(s) deleted due to missingness )
( 313156 observation(s) deleted due to perfect classification )
Number of Fisher Scoring Iterations: 25
Average individual fixed effect= -32515929477082
Part 2: Monadic Models
Monadic Models Methods
The Monadic Models draw on data from the First Use of Violent Force Dataset (Caprioli and Trumbore 2006). This is a much smaller dataset both in time and scope. The data examines all MIDS from 1980 to 2002. Because the data size is already reduced, I do not use the rare events or weighted OLS models. I use a panel linear model, a panel GLM model, and a standard logit model for each.
Authors cited
Complete citations are included in the manuscript and will be updated here as this draft is completed.