A data set on civilian casualties in civil war; illustrates count regressions
Source:R/hks2013.R
hks2013.RdHultman, Kathman, and Shannon (2013) data set used in McCaskey and Rainey (2015) to illustrate count regression models, such as Poisson, negative binomial, and their zero-inflated and hurdle variants. These are the data to reproduce the negative binomial regression coefficients for Model 1 in Table 1 on p. 884 of Hultman, Kathman, and Shannon (2013).
Format
A country-month data frame.
locationthe country
yearthe year
monthe month
osvAllthe number of civilians killed in a conflict month by any combatant faction
troopLagthe number of UN military troops committed to a country in conflict during a given month; lagged one month
policeLagthe number of UN police committed to a country in conflict during a given month; lagged one month
militaryobserversLagthe number of UN observers committed to a country in conflict during a given month; lagged one month
brv_AllLagthe number of all battle deaths; lagged one month
osvAllLagDumwhether any civilians were killed by any combatant faction; lagged one month
incompa dichotomous variable that uses the UCDP/PRIO delineation of civil wars fought over territorial (0) or government (1) control.
epdurationthe number of months since the beginning of a conflict episode
lntpopthe country's logged population size
For further details, see Hultman, Kathman, and Shannon (2013) pp. 883-884 and Kathman (2013).
References
Hultman, Lisa, Jacob Kathman, and Megan Shannon. 2013. "United Nations Peacekeeping and Civilian Protection in Civil War." American Journal of Political Science 57(4): 875–91. doi:10.1111/ajps.12036 .
Hultman, Lisa, Jacob Kathman, and Megan Shannon. 2013. "Replication data for: United Nations Peacekeeping and Civilian Protection in Civil War." Harvard Dataverse, V3. doi:10.7910/DVN/6EBCGA .
Kathman, Jacob D. 2013. "United Nations Peacekeeping Personnel Commitments, 1990–2011." Conflict Management and Peace Science 30(5): 532–49. doi:10.1177/0738894213491180 .
McCaskey, Kelly, and Carlisle Rainey. 2015. "Substantive Importance and the Veil of Statistical Significance." Statistics, Politics and Policy 6(1–2). doi:10.1515/spp-2015-0001 .
Examples
# a simple example
# load data in a way that mirrors read.csv(), etc.
hks <- crdata::hks2013
# estimate models
f <- osvAll ~ troopLag + policeLag + militaryobserversLag +
brv_AllLag + osvAllLagDum + incomp + epduration +
lntpop
# replicates model 1 in table 1 on p. 884 of HKS
# m2 <- MASS::glm.nb(f, data = hks2013,
# init.theta = 5,
# control = glm.control(epsilon = 1e-12,
# maxit = 2500,
# trace = FALSE))
# summary(m2) # compare to p. 884
# a poisson alternative to the above
m1 <- glm(f, data = hks2013, family = "poisson")
summary(m1)
#>
#> Call:
#> glm(formula = f, family = "poisson", data = hks2013)
#>
#> Coefficients:
#> Estimate Std. Error z value Pr(>|z|)
#> (Intercept) -3.579e+00 4.259e-02 -84.04 <2e-16 ***
#> troopLag -1.697e-01 1.758e-03 -96.52 <2e-16 ***
#> policeLag -3.272e+00 2.448e-02 -133.70 <2e-16 ***
#> militaryobserversLag 8.100e+00 1.206e-02 671.54 <2e-16 ***
#> brv_AllLag 5.606e-04 8.476e-06 66.14 <2e-16 ***
#> osvAllLagDum 2.911e-01 4.740e-03 61.42 <2e-16 ***
#> incomp 3.486e+00 1.800e-02 193.69 <2e-16 ***
#> epduration -2.223e-02 7.940e-05 -279.98 <2e-16 ***
#> lntpop 1.894e-01 2.076e-03 91.24 <2e-16 ***
#> ---
#> Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
#>
#> (Dispersion parameter for poisson family taken to be 1)
#>
#> Null deviance: 2840157 on 3745 degrees of freedom
#> Residual deviance: 2134801 on 3737 degrees of freedom
#> AIC: 2139137
#>
#> Number of Fisher Scoring iterations: 11
#>