Regression Tables

November 3, 2026

What’s in a Regression Table?

Load Data


library(peacesciencer)
library(dplyr)

conflict_df <- create_stateyears(system = 'gw') |>
  filter(year %in% c(1946:1999)) |>
  add_ucdp_acd(type=c("intrastate"), only_wars = FALSE) |>
  add_democracy() |>
  add_creg_fractionalization() |>
  add_sim_gdp_pop() |>
  add_rugged_terrain() |>
  mutate(gdppc = log(mrgdppc), # log GDP per capita
         pop = log(pwtpop))    # log population

glimpse(conflict_df)

Run Model


conflict_model <- glm(ucdponset ~ ethfrac + relfrac + v2x_polyarchy + 
                        rugged + gdppc + pop,
                  data= conflict_df,
                  family = binomial(link="logit"))

View with summary()

summary(conflict_model)

Call:
glm(formula = ucdponset ~ ethfrac + relfrac + v2x_polyarchy + 
    rugged + gdppc + pop, family = binomial(link = "logit"), 
    data = conflict_df)

Coefficients:
              Estimate Std. Error z value Pr(>|z|)    
(Intercept)   -1.62860    0.99050  -1.644  0.10013    
ethfrac        0.71013    0.37159   1.911  0.05600 .  
relfrac       -0.24805    0.40739  -0.609  0.54260    
v2x_polyarchy -0.73674    0.50941  -1.446  0.14810    
rugged         0.08712    0.07567   1.151  0.24959    
gdppc         -0.37570    0.11969  -3.139  0.00169 ** 
pop            0.29367    0.06866   4.277 1.89e-05 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

(Dispersion parameter for binomial family taken to be 1)

    Null deviance: 1205.0  on 6128  degrees of freedom
Residual deviance: 1148.6  on 6122  degrees of freedom
  (1495 observations deleted due to missingness)
AIC: 1162.6

Number of Fisher Scoring iterations: 7

View results with broom

library(broom)

tidy_model <- conflict_model |>
  tidy(conf.int = TRUE) |>
  mutate_if(is.numeric, round, 5)

tidy_model
# A tibble: 7 × 7
  term          estimate std.error statistic p.value conf.low conf.high
  <chr>            <dbl>     <dbl>     <dbl>   <dbl>    <dbl>     <dbl>
1 (Intercept)    -1.63      0.990     -1.64  0.100   -3.56        0.326
2 ethfrac         0.710     0.372      1.91  0.056   -0.00629     1.45 
3 relfrac        -0.248     0.407     -0.609 0.543   -1.05        0.547
4 v2x_polyarchy  -0.737     0.509     -1.45  0.148   -1.75        0.252
5 rugged          0.0871    0.0757     1.15  0.250   -0.0680      0.229
6 gdppc          -0.376     0.120     -3.14  0.00169 -0.615      -0.146
7 pop             0.294     0.0687     4.28  0.00002  0.160       0.430

How close are our results to F&L?


Discuss with a neighbor…

tidy_model
# A tibble: 7 × 7
  term          estimate std.error statistic p.value conf.low conf.high
  <chr>            <dbl>     <dbl>     <dbl>   <dbl>    <dbl>     <dbl>
1 (Intercept)    -1.63      0.990     -1.64  0.100   -3.56        0.326
2 ethfrac         0.710     0.372      1.91  0.056   -0.00629     1.45 
3 relfrac        -0.248     0.407     -0.609 0.543   -1.05        0.547
4 v2x_polyarchy  -0.737     0.509     -1.45  0.148   -1.75        0.252
5 rugged          0.0871    0.0757     1.15  0.250   -0.0680      0.229
6 gdppc          -0.376     0.120     -3.14  0.00169 -0.615      -0.146
7 pop             0.294     0.0687     4.28  0.00002  0.160       0.430

Could we get closer to F&L?


  • How close are our data to F&L’s?
  • Got to the peacesciencer documentation
  • Could we change something to better approximate their results?
  • Are there predictors in the peacesciencer package that were not available to F&L?

Regression Tables with modelsummary


  • Oftentimes we want to show multiple models at once (like F&L)
  • We want to compare across them and see which is the best model
  • How can we do that?
  • There are many ways to do this in R
  • We will use the modelsummary package

Run Multiple Models


ethnicity <- glm(ucdponset ~ ethfrac + relfrac + gdppc + pop, # store each model in an object
                  data = conflict_df,
                  family = "binomial")

democracy <- glm(ucdponset ~ v2x_polyarchy + gdppc +  pop,
                  data = conflict_df,
                  family = "binomial")

terrain <- glm(ucdponset ~ rugged + gdppc + pop ,
                  data = conflict_df,
                  family = "binomial")

full_model <- glm(ucdponset ~ ethfrac + relfrac + v2x_polyarchy + rugged +
                        gdppc + pop,
                  data = conflict_df,
                  family = "binomial")

Prep Data for Display


models <- list("Ethnicity" = ethnicity,  # store list of models in an object
               "Democracy" = democracy, 
               "Terrain" = terrain, 
               "Full Model" = full_model)

coef_map <- c("ethfrac" = "Ethnic Frac",  # map coefficients
        "relfrac" = "Religions Frac",     #(change names and order)
        "v2x_polyarchy" = "Polyarchy",
        "rugged" = "Terrain",
        "gdppc" = "Per capita GDP",
        "pop" = "Population",
        "(Intercept)" = "Intercept")

caption = "Table 1: Predictors of Conflict Onset" # store caption
reference = "See appendix for data sources."      # store reference notes

Display the Models

Code 
library(modelsummary)

modelsummary(models,                      # display the table
             stars = TRUE,                # include stars for significance
             gof_map = c("nobs"),         # goodness of fit stats to include   
             coef_map = coef_map,         # coefficient mapping
             title = caption,             # title
             notes = reference)           # source note
Table 1: Predictors of Conflict Onset
Ethnicity Democracy Terrain Full Model
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001
See appendix for data sources.
Ethnic Frac 0.747* 0.710+
(0.364) (0.372)
Religions Frac -0.373 -0.248
(0.406) (0.407)
Polyarchy -0.349 -0.737
(0.436) (0.509)
Terrain 0.036 0.087
(0.075) (0.076)
Per capita GDP -0.478*** -0.497*** -0.539*** -0.376**
(0.104) (0.105) (0.092) (0.120)
Population 0.290*** 0.302*** 0.303*** 0.294***
(0.068) (0.051) (0.050) (0.069)
Intercept -0.887 -0.425 -0.254 -1.629
(0.898) (0.791) (0.758) (0.991)
Num.Obs. 6364 6955 6840 6129

Your Turn!


  • Got to the peacesciencer documentation
  • How close are our data to F&L’s?
  • Could we change something to better approximate their results?
  • Run multiple models using different predictors
  • Display the models using modelsummary
  • Try to get as close to F&L as you can!

Coefficient Plots


This doesn’t look that great…

Show the code 
modelsummary(conflict_model, 
             stars = TRUE,  
             gof_map = c("nobs"),
             coef_map = coef_map,
             title = caption, 
             notes = reference)
Table 1: Predictors of Conflict Onset
(1)
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001
See appendix for data sources.
Ethnic Frac 0.710+
(0.372)
Religions Frac -0.248
(0.407)
Polyarchy -0.737
(0.509)
Terrain 0.087
(0.076)
Per capita GDP -0.376**
(0.120)
Population 0.294***
(0.069)
Intercept -1.629
(0.991)
Num.Obs. 6129


So we can use ggplot to make a coefficient plot instead…

Show the code 
library(ggplot2)

modelplot(conflict_model, 
          coef_map = rev(coef_map), # rev() reverses list order
          coef_omit = "Intercept", 
          color = "blue") + # use plus to add customizations like any ggplot object
  geom_vline(xintercept = 0, color = "red", linetype = "dashed", linewidth = .75) + # red 0 line
  labs(
    title = "Figure 1: Predictors of Conflict Onset",
    caption = "See appendix for data sources."
  )

Your Turn!

  • Take one of your models
  • Use modelplot to create a coefficient plot of it
  • Customize the plot to your liking
  • Interpret the results
  • Discuss advantages of coefficient plots with a neighbor