Analysis of a Conjoint Survey Experiment

dplyr

data science

Published

April 26, 2026

Below is the analysis of a conjoint survey experiment that I performed for Professor Mara Revkin at Duke University. I am not able to include the details of the survey experiment such as the text of the questions at this time.

suppressPackageStartupMessages(library(lmtest))
suppressPackageStartupMessages(library(sandwich))
suppressPackageStartupMessages(library(tidyverse))

In the questionnaire, this analysis covers chapter 10 questions 74-78. There 3 iterations of the experimental treatment per respondent. The relevant variables, where x is the iteration number 1-3:

I_x_Q_75 – iteration x, randomization 1
I_x_Q_76 – iteration x, randomization 2
I_x_Q_77 – iteration x, randomization 3
I_x_Q_78 – iteration x, randomization 4
I_x_Q6_4 – iteration x, response 1
I_x_Q6_5_A – iteration x, response 2

Data Wrangling

df <- read_csv("data/feb9data.csv")

For response variable 2, Q6_5_A, I am going to:

convert -1 to NA
convert 2. No to 0
convert 1. Yes to 1

I use sapply() to convert the responses for this variable.

convert_q6_5a <- function(x) {
  if (x == "2. No") return(0)
  else if (x == "1. Yes") return(1)
  else return(NA)
}

df$I_1_Q6_5_A <- unname(sapply(df$I_1_Q6_5_A, convert_q6_5a))
df$I_2_Q6_5_A <- unname(sapply(df$I_2_Q6_5_A, convert_q6_5a))
df$I_3_Q6_5_A <- unname(sapply(df$I_3_Q6_5_A, convert_q6_5a))

I’m going to select the subset of columns that I want to actually do analysis on, using dplyr’s ends_with() select helper:

df_sub <- df |> 
  select(SbjNum,
         ends_with("_Q_75"),
         ends_with("_Q_76"),
         ends_with("_Q_77"),
         ends_with("_Q_78"),
         ends_with("_Q6_4"),
         ends_with("_Q6_5_A"))

Now we work on reshaping the data using gather(). I conduct a few QA checks to ensure that the reshaping did not result in any data loss.

gathered_vars <- df_sub |> 
  select(SbjNum,
         ends_with("_Q_75"),
         ends_with("_Q_76"),
         ends_with("_Q_77"),
         ends_with("_Q_78")) |> 
  gather(variable, value, -SbjNum) 

iterations <- as.numeric(
  str_match(gathered_vars$variable, "I_(\\d)_Q_7.")[, 2]
)
q_num <- str_match(gathered_vars$variable, "I_\\d_(Q_\\d{2})")[, 2]

Let’s look at iterations and q_num to make sure they have the right number of values and the correct distributions:

length(iterations) == nrow(gathered_vars)

[1] TRUE

length(q_num) == nrow(gathered_vars)

[1] TRUE

table(iterations)

iterations
   1    2    3 
3004 3004 3004

table(q_num)

q_num
Q_75 Q_76 Q_77 Q_78 
2253 2253 2253 2253

Ok, they all have the correct number of items all equally distributed. I now create a new data frame with the reshaped data.

reshaped_vars <- cbind(gathered_vars, iterations, q_num) |> 
  select(-variable) |> 
  spread(q_num, value) |> 
  rename(iteration = iterations)

Quality assurance: Randomly select 5 subject IDs and confirm that the values are the same between the original data frame and the gathered data:

qa_subjects <- sample(gathered_vars$SbjNum, 5)

for (subject in qa_subjects) {
  print(paste("Subject:", subject))
  df_subject_original <- filter(df_sub, SbjNum == subject)
  df_subject_gathered <- filter(reshaped_vars, SbjNum == subject)
  
  for (iter in 1:3) {
    # Q_75
    gathered_75 <- df_subject_gathered |> 
      filter(iteration == iter) |> 
      select(Q_75) |> 
      pull()
    print(paste(
      "Iteration", 
      iter,
      "Q_75:", 
      df_subject_original[[paste0("I_", iter, "_Q_75")]] == gathered_75
    ))
    
    # Q_76
    gathered_76 <- df_subject_gathered |> 
      filter(iteration == iter) |> 
      select(Q_76) |> 
      pull()
    print(paste(
      "Iteration",
      iter,
      "Q_76:", 
      df_subject_original[[paste0("I_", iter, "_Q_76")]] == gathered_76
    ))
    
    # Q_77
    gathered_77 <- df_subject_gathered |> 
      filter(iteration == iter) |> 
      select(Q_77) |> 
      pull()
    print(paste(
      "Iteration", 
      iter,
      "Q_77:", 
      df_subject_original[[paste0("I_", iter, "_Q_77")]] == gathered_77
    ))
    
    # Q_78
    gathered_78 <- df_subject_gathered |> 
      filter(iteration == iter) |> 
      select(Q_78) |> 
      pull()
    print(paste(
      "Iteration",
      iter,
      "Q_78:", 
      df_subject_original[[paste0("I_", iter, "_Q_78")]] == gathered_78
    ))
  }
}

[1] "Subject: 214698768"
[1] "Iteration 1 Q_75: TRUE"
[1] "Iteration 1 Q_76: TRUE"
[1] "Iteration 1 Q_77: TRUE"
[1] "Iteration 1 Q_78: TRUE"
[1] "Iteration 2 Q_75: TRUE"
[1] "Iteration 2 Q_76: TRUE"
[1] "Iteration 2 Q_77: TRUE"
[1] "Iteration 2 Q_78: TRUE"
[1] "Iteration 3 Q_75: TRUE"
[1] "Iteration 3 Q_76: TRUE"
[1] "Iteration 3 Q_77: TRUE"
[1] "Iteration 3 Q_78: TRUE"
[1] "Subject: 214645315"
[1] "Iteration 1 Q_75: TRUE"
[1] "Iteration 1 Q_76: TRUE"
[1] "Iteration 1 Q_77: TRUE"
[1] "Iteration 1 Q_78: TRUE"
[1] "Iteration 2 Q_75: TRUE"
[1] "Iteration 2 Q_76: TRUE"
[1] "Iteration 2 Q_77: TRUE"
[1] "Iteration 2 Q_78: TRUE"
[1] "Iteration 3 Q_75: TRUE"
[1] "Iteration 3 Q_76: TRUE"
[1] "Iteration 3 Q_77: TRUE"
[1] "Iteration 3 Q_78: TRUE"
[1] "Subject: 214914473"
[1] "Iteration 1 Q_75: TRUE"
[1] "Iteration 1 Q_76: TRUE"
[1] "Iteration 1 Q_77: TRUE"
[1] "Iteration 1 Q_78: TRUE"
[1] "Iteration 2 Q_75: TRUE"
[1] "Iteration 2 Q_76: TRUE"
[1] "Iteration 2 Q_77: TRUE"
[1] "Iteration 2 Q_78: TRUE"
[1] "Iteration 3 Q_75: TRUE"
[1] "Iteration 3 Q_76: TRUE"
[1] "Iteration 3 Q_77: TRUE"
[1] "Iteration 3 Q_78: TRUE"
[1] "Subject: 215151139"
[1] "Iteration 1 Q_75: TRUE"
[1] "Iteration 1 Q_76: TRUE"
[1] "Iteration 1 Q_77: TRUE"
[1] "Iteration 1 Q_78: TRUE"
[1] "Iteration 2 Q_75: TRUE"
[1] "Iteration 2 Q_76: TRUE"
[1] "Iteration 2 Q_77: TRUE"
[1] "Iteration 2 Q_78: TRUE"
[1] "Iteration 3 Q_75: TRUE"
[1] "Iteration 3 Q_76: TRUE"
[1] "Iteration 3 Q_77: TRUE"
[1] "Iteration 3 Q_78: TRUE"
[1] "Subject: 214992513"
[1] "Iteration 1 Q_75: TRUE"
[1] "Iteration 1 Q_76: TRUE"
[1] "Iteration 1 Q_77: TRUE"
[1] "Iteration 1 Q_78: TRUE"
[1] "Iteration 2 Q_75: TRUE"
[1] "Iteration 2 Q_76: TRUE"
[1] "Iteration 2 Q_77: TRUE"
[1] "Iteration 2 Q_78: TRUE"
[1] "Iteration 3 Q_75: TRUE"
[1] "Iteration 3 Q_76: TRUE"
[1] "Iteration 3 Q_77: TRUE"
[1] "Iteration 3 Q_78: TRUE"

That all looks good, I feel confident that the reshaping didn’t mess up any of the data points.

Now I need to do the same kind of reshaping on the response variables.

gathered_responses <- df_sub |> 
  select(SbjNum,
         matches("I_\\d_Q6_\\d*")
         ) |> 
  gather(variable, value, -SbjNum) 

iterations <- as.numeric(
  str_match(gathered_responses$variable, "I_(\\d)_Q6")[, 2]
)
q_num <- str_match(gathered_responses$variable, "I_\\d_(Q6.+)")[, 2]

Ensure correct number of items:

length(iterations) == nrow(gathered_responses)

[1] TRUE

length(q_num) == nrow(gathered_responses)

[1] TRUE

table(iterations)

iterations
   1    2    3 
1502 1502 1502

table(q_num)

q_num
  Q6_4 Q6_5_A 
  2253   2253

That all looks good.

reshaped_responses <- cbind(gathered_responses, iterations, q_num) |> 
  select(-variable) |> 
  spread(q_num, value) |> 
  rename(iteration = iterations)

Repeat the same kind of QA for the response variables:

qa_subjects <- sample(reshaped_responses$SbjNum, 5)

for (subject in qa_subjects) {
  print(paste("Subject:", subject))
  df_subject_original <- filter(df_sub, SbjNum == subject)
  df_subject_reshaped <- filter(reshaped_responses, SbjNum == subject)
  
  for (iter in 1:3) {
    # Q6_4
    reshaped_6_4 <- df_subject_reshaped |> 
      filter(iteration == iter) |> 
      select(Q6_4) |> 
      pull()
    original_response <- df_subject_original[[paste0("I_", iter, "_Q6_4")]]
    if (is.na(original_response) && is.na(reshaped_64)) {
      print(paste("Iteration", iter, "Q6_4:", TRUE))
    } else {
      print(paste(
        "Iteration",
        iter,
        "Q6_4:",
        original_response == reshaped_6_4
      ))
    }
    
    # Q6_5_A
    reshaped_6_5_A <- df_subject_reshaped |> 
      filter(iteration == iter) |> 
      select(Q6_5_A) |> 
      pull()
    original_response <- df_subject_original[[paste0("I_", iter, "_Q6_5_A")]]
    if (is.na(original_response) && is.na(reshaped_6_5_A)) {
      print(paste("Iteration", iter, "Q6_5_A:", TRUE))
    } else {
      print(paste(
        "Iteration",
        iter, 
        "Q6_5_A:", 
        original_response == reshaped_6_5_A
      ))
    }
  }
}

[1] "Subject: 214956721"
[1] "Iteration 1 Q6_4: TRUE"
[1] "Iteration 1 Q6_5_A: TRUE"
[1] "Iteration 2 Q6_4: TRUE"
[1] "Iteration 2 Q6_5_A: TRUE"
[1] "Iteration 3 Q6_4: TRUE"
[1] "Iteration 3 Q6_5_A: TRUE"
[1] "Subject: 214630669"
[1] "Iteration 1 Q6_4: TRUE"
[1] "Iteration 1 Q6_5_A: TRUE"
[1] "Iteration 2 Q6_4: TRUE"
[1] "Iteration 2 Q6_5_A: TRUE"
[1] "Iteration 3 Q6_4: TRUE"
[1] "Iteration 3 Q6_5_A: TRUE"
[1] "Subject: 214697633"
[1] "Iteration 1 Q6_4: TRUE"
[1] "Iteration 1 Q6_5_A: TRUE"
[1] "Iteration 2 Q6_4: TRUE"
[1] "Iteration 2 Q6_5_A: TRUE"
[1] "Iteration 3 Q6_4: TRUE"
[1] "Iteration 3 Q6_5_A: TRUE"
[1] "Subject: 214647860"
[1] "Iteration 1 Q6_4: TRUE"
[1] "Iteration 1 Q6_5_A: TRUE"
[1] "Iteration 2 Q6_4: TRUE"
[1] "Iteration 2 Q6_5_A: TRUE"
[1] "Iteration 3 Q6_4: TRUE"
[1] "Iteration 3 Q6_5_A: TRUE"
[1] "Subject: 215029665"
[1] "Iteration 1 Q6_4: TRUE"
[1] "Iteration 1 Q6_5_A: TRUE"
[1] "Iteration 2 Q6_4: TRUE"
[1] "Iteration 2 Q6_5_A: TRUE"
[1] "Iteration 3 Q6_4: TRUE"
[1] "Iteration 3 Q6_5_A: TRUE"

All looks good with the response reshaping.

Now we can join the two data frames back together:

joined <- left_join(
  reshaped_vars,
  reshaped_responses,
  by = c("SbjNum", "iteration")
)

I will now convert the randomization variables and Q6_4 to factor, and Q6_5_A to numeric.

joined <- joined |> 
  mutate(
    across(contains("Q_7"), ~ factor(.x)),
    Q6_4 =  factor(Q6_4,
                   levels = c(
                     "1. Fully sufficient",
                     "2. Somewhat sufficient",
                     "3. Somewhat insufficient but better than nothing",
                     "4. Totally insufficient"
                   ),
                   labels = c(
                     "1. Fully sufficient",
                     "2. Somewhat sufficient",
                     "3. Somewhat insufficient",
                     "4. Totally insufficient"
                   ),
                   ordered = TRUE),
    Q6_5_A = as.numeric(Q6_5_A)
 )

I like to make a “final copy” of the data after I’m done.

cleaned_data <- joined

Analysis

I think I can now run a couple basic analyses. First let’s look at the distribution of response variables:

table(cleaned_data$Q6_4, useNA = "always")


     1. Fully sufficient   2. Somewhat sufficient 3. Somewhat insufficient 
                     473                      525                      443 
 4. Totally insufficient                     <NA> 
                     810                        2

A plurality of respondents answered “4. Totally insufficient”, with the other options quite a bit lower.

table(cleaned_data$Q6_5_A, useNA = "always")


   0    1 <NA> 
 680  917  656

Yes has more responses than No, but there are a large number of NAs.

Ordered logit without respondent effects

I will now look at an ordered logit regression for Q6_4. This model does not control for subject.

Please note, due to the length of the response values the tables below are extremely wide. You can scroll side to side to show the point estimates and other relevant statistics.

q6_4_logit <- MASS::polr(Q6_4 ~ Q_75 + Q_76 + Q_77 + Q_78, 
                         data = cleaned_data, 
                         method = "logistic",
                         Hess = TRUE)

# Compute Odds Ratios and Confidence Intervals
exp(cbind(OR = coef(q6_4_logit), confint(q6_4_logit)))

                                                                                                                                    OR
Q_75We would like to apologize for the death of your loved one.                                                              0.8823071
Q_75We would like to express our sympathy for the death of your loved one.                                                   0.9535238
Q_76We can offer you a one-time payment of 3 million Iraqi Dinars as compensation for your loss.                             0.7707824
Q_76We can offer you a one-time payment of 30 million Iraqi Dinars as compensation for your loss.                            0.3205635
Q_77We would like to assure you that we have incorporated the lessons learned from your case into our training curriculum to 0.8578855
Q_77We would like to assure you that we thoroughly investigated your case and found no evidence of wrongdoing by Iraqi force 0.8303470
Q_78The Iraqi government has agreed to build a public memorial to commemorate the victims of Daesh and the battle for Mosul  0.7134516
Q_78The Iraqi government has agreed to build three new schools in Mosul to support the city’s recovery from the losses suf   0.6427085
                                                                                                                                 2.5 %
Q_75We would like to apologize for the death of your loved one.                                                              0.7361427
Q_75We would like to express our sympathy for the death of your loved one.                                                   0.7896712
Q_76We can offer you a one-time payment of 3 million Iraqi Dinars as compensation for your loss.                             0.6365587
Q_76We can offer you a one-time payment of 30 million Iraqi Dinars as compensation for your loss.                            0.2658655
Q_77We would like to assure you that we have incorporated the lessons learned from your case into our training curriculum to 0.7127676
Q_77We would like to assure you that we thoroughly investigated your case and found no evidence of wrongdoing by Iraqi force 0.6896368
Q_78The Iraqi government has agreed to build a public memorial to commemorate the victims of Daesh and the battle for Mosul  0.5939732
Q_78The Iraqi government has agreed to build three new schools in Mosul to support the city’s recovery from the losses suf   0.5330191
                                                                                                                                97.5 %
Q_75We would like to apologize for the death of your loved one.                                                              1.0573268
Q_75We would like to express our sympathy for the death of your loved one.                                                   1.1513703
Q_76We can offer you a one-time payment of 3 million Iraqi Dinars as compensation for your loss.                             0.9330271
Q_76We can offer you a one-time payment of 30 million Iraqi Dinars as compensation for your loss.                            0.3860622
Q_77We would like to assure you that we have incorporated the lessons learned from your case into our training curriculum to 1.0323678
Q_77We would like to assure you that we thoroughly investigated your case and found no evidence of wrongdoing by Iraqi force 0.9995685
Q_78The Iraqi government has agreed to build a public memorial to commemorate the victims of Daesh and the battle for Mosul  0.8566594
Q_78The Iraqi government has agreed to build three new schools in Mosul to support the city’s recovery from the losses suf   0.7746367

It looks like Q75, the messaging around the loss, doesn’t contain significant effects across feature levels at the 0.05 level.

Note that Q6_4 is ordered such that lower values indicate greater sufficiency. I can reverse that if we want. So in looking at the data, an odds ratio below 1 indicates a higher degree of sufficiency.

There are a number of feature levels where the confidence interval does not contain 1:

Q76: One-time payment of 3 million Dinars: OR 0.771
Q76: One-time payment of 30 million Dinars: OR 0.321
Q77: No evidence of wrongdoing: 0.830
Q78: Public memorial: 0.713
Q78: New schools: 0.643

I can cluster standard errors by respondent using the sandwich package:

coeftest(q6_4_logit, vcov = vcovCL(q6_4_logit, cluster = ~ SbjNum))


t test of coefficients:

                                                                                                                              Estimate
Q_75We would like to apologize for the death of your loved one.                                                              -0.125215
Q_75We would like to express our sympathy for the death of your loved one.                                                   -0.047591
Q_76We can offer you a one-time payment of 3 million Iraqi Dinars as compensation for your loss.                             -0.260349
Q_76We can offer you a one-time payment of 30 million Iraqi Dinars as compensation for your loss.                            -1.137675
Q_77We would like to assure you that we have incorporated the lessons learned from your case into our training curriculum to -0.153285
Q_77We would like to assure you that we thoroughly investigated your case and found no evidence of wrongdoing by Iraqi force -0.185912
Q_78The Iraqi government has agreed to build a public memorial to commemorate the victims of Daesh and the battle for Mosul  -0.337641
Q_78The Iraqi government has agreed to build three new schools in Mosul to support the city’s recovery from the losses suf   -0.442064
                                                                                                                             Std. Error
Q_75We would like to apologize for the death of your loved one.                                                                0.092036
Q_75We would like to express our sympathy for the death of your loved one.                                                     0.098503
Q_76We can offer you a one-time payment of 3 million Iraqi Dinars as compensation for your loss.                               0.104934
Q_76We can offer you a one-time payment of 30 million Iraqi Dinars as compensation for your loss.                              0.108773
Q_77We would like to assure you that we have incorporated the lessons learned from your case into our training curriculum to   0.094921
Q_77We would like to assure you that we thoroughly investigated your case and found no evidence of wrongdoing by Iraqi force   0.096449
Q_78The Iraqi government has agreed to build a public memorial to commemorate the victims of Daesh and the battle for Mosul    0.094216
Q_78The Iraqi government has agreed to build three new schools in Mosul to support the city’s recovery from the losses suf     0.099666
                                                                                                                              t value
Q_75We would like to apologize for the death of your loved one.                                                               -1.3605
Q_75We would like to express our sympathy for the death of your loved one.                                                    -0.4831
Q_76We can offer you a one-time payment of 3 million Iraqi Dinars as compensation for your loss.                              -2.4811
Q_76We can offer you a one-time payment of 30 million Iraqi Dinars as compensation for your loss.                            -10.4591
Q_77We would like to assure you that we have incorporated the lessons learned from your case into our training curriculum to  -1.6149
Q_77We would like to assure you that we thoroughly investigated your case and found no evidence of wrongdoing by Iraqi force  -1.9276
Q_78The Iraqi government has agreed to build a public memorial to commemorate the victims of Daesh and the battle for Mosul   -3.5837
Q_78The Iraqi government has agreed to build three new schools in Mosul to support the city’s recovery from the losses suf    -4.4355
                                                                                                                              Pr(>|t|)
Q_75We would like to apologize for the death of your loved one.                                                               0.173809
Q_75We would like to express our sympathy for the death of your loved one.                                                    0.629041
Q_76We can offer you a one-time payment of 3 million Iraqi Dinars as compensation for your loss.                              0.013172
Q_76We can offer you a one-time payment of 30 million Iraqi Dinars as compensation for your loss.                            < 2.2e-16
Q_77We would like to assure you that we have incorporated the lessons learned from your case into our training curriculum to  0.106481
Q_77We would like to assure you that we thoroughly investigated your case and found no evidence of wrongdoing by Iraqi force  0.054036
Q_78The Iraqi government has agreed to build a public memorial to commemorate the victims of Daesh and the battle for Mosul   0.000346
Q_78The Iraqi government has agreed to build three new schools in Mosul to support the city’s recovery from the losses suf   9.632e-06
                                                                                                                                
Q_75We would like to apologize for the death of your loved one.                                                                 
Q_75We would like to express our sympathy for the death of your loved one.                                                      
Q_76We can offer you a one-time payment of 3 million Iraqi Dinars as compensation for your loss.                             *  
Q_76We can offer you a one-time payment of 30 million Iraqi Dinars as compensation for your loss.                            ***
Q_77We would like to assure you that we have incorporated the lessons learned from your case into our training curriculum to    
Q_77We would like to assure you that we thoroughly investigated your case and found no evidence of wrongdoing by Iraqi force .  
Q_78The Iraqi government has agreed to build a public memorial to commemorate the victims of Daesh and the battle for Mosul  ***
Q_78The Iraqi government has agreed to build three new schools in Mosul to support the city’s recovery from the losses suf   ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

With the clustered standard errors, we still see significant results at the 0.05 level for:

Q76: 3 million Dinars
Q76: 30 million Dinars
Q78: Public memorial
Q78: New schools

Note that Q77: no evidence of wrongdoing is close to p < 0.05.

Binary logit for Q6_5_A

Would the respondent accept the offer of compensation?

First, exclude randomizations that do not include compensation, and re-label the factor

compensation_rows <- cleaned_data |> 
  filter(Q_76 != "Unfortunately, we cannot offer you any monetary compensation for your loss.") |> 
  mutate(Q_76 = factor(
    Q_76,
    levels = c(
      "We can offer you a one-time payment of 3 million Iraqi Dinars as compensation for your loss.",
      "We can offer you a one-time payment of 30 million Iraqi Dinars as compensation for your loss."
      )
    )
    )

q6_5_a_logit <- glm(Q6_5_A ~ Q_75 + Q_76 + Q_77 + Q_78, 
                    data = compensation_rows,
                    family = binomial(link = "logit"))
summary(q6_5_a_logit)


Call:
glm(formula = Q6_5_A ~ Q_75 + Q_76 + Q_77 + Q_78, family = binomial(link = "logit"), 
    data = compensation_rows)

Coefficients:
                                                                                                                             Estimate
(Intercept)                                                                                                                  -0.54051
Q_75We would like to apologize for the death of your loved one.                                                               0.13688
Q_75We would like to express our sympathy for the death of your loved one.                                                    0.04118
Q_76We can offer you a one-time payment of 30 million Iraqi Dinars as compensation for your loss.                             1.25561
Q_77We would like to assure you that we have incorporated the lessons learned from your case into our training curriculum to -0.18102
Q_77We would like to assure you that we thoroughly investigated your case and found no evidence of wrongdoing by Iraqi force  0.03259
Q_78The Iraqi government has agreed to build a public memorial to commemorate the victims of Daesh and the battle for Mosul   0.16550
Q_78The Iraqi government has agreed to build three new schools in Mosul to support the city’s recovery from the losses suf    0.04471
                                                                                                                             Std. Error
(Intercept)                                                                                                                     0.15403
Q_75We would like to apologize for the death of your loved one.                                                                 0.13065
Q_75We would like to express our sympathy for the death of your loved one.                                                      0.13543
Q_76We can offer you a one-time payment of 30 million Iraqi Dinars as compensation for your loss.                               0.10915
Q_77We would like to assure you that we have incorporated the lessons learned from your case into our training curriculum to    0.13373
Q_77We would like to assure you that we thoroughly investigated your case and found no evidence of wrongdoing by Iraqi force    0.13382
Q_78The Iraqi government has agreed to build a public memorial to commemorate the victims of Daesh and the battle for Mosul     0.13157
Q_78The Iraqi government has agreed to build three new schools in Mosul to support the city’s recovery from the losses suf      0.13388
                                                                                                                             z value
(Intercept)                                                                                                                   -3.509
Q_75We would like to apologize for the death of your loved one.                                                                1.048
Q_75We would like to express our sympathy for the death of your loved one.                                                     0.304
Q_76We can offer you a one-time payment of 30 million Iraqi Dinars as compensation for your loss.                             11.504
Q_77We would like to assure you that we have incorporated the lessons learned from your case into our training curriculum to  -1.354
Q_77We would like to assure you that we thoroughly investigated your case and found no evidence of wrongdoing by Iraqi force   0.244
Q_78The Iraqi government has agreed to build a public memorial to commemorate the victims of Daesh and the battle for Mosul    1.258
Q_78The Iraqi government has agreed to build three new schools in Mosul to support the city’s recovery from the losses suf     0.334
                                                                                                                             Pr(>|z|)
(Intercept)                                                                                                                   0.00045
Q_75We would like to apologize for the death of your loved one.                                                               0.29479
Q_75We would like to express our sympathy for the death of your loved one.                                                    0.76105
Q_76We can offer you a one-time payment of 30 million Iraqi Dinars as compensation for your loss.                             < 2e-16
Q_77We would like to assure you that we have incorporated the lessons learned from your case into our training curriculum to  0.17586
Q_77We would like to assure you that we thoroughly investigated your case and found no evidence of wrongdoing by Iraqi force  0.80758
Q_78The Iraqi government has agreed to build a public memorial to commemorate the victims of Daesh and the battle for Mosul   0.20842
Q_78The Iraqi government has agreed to build three new schools in Mosul to support the city’s recovery from the losses suf    0.73842
                                                                                                                                
(Intercept)                                                                                                                  ***
Q_75We would like to apologize for the death of your loved one.                                                                 
Q_75We would like to express our sympathy for the death of your loved one.                                                      
Q_76We can offer you a one-time payment of 30 million Iraqi Dinars as compensation for your loss.                            ***
Q_77We would like to assure you that we have incorporated the lessons learned from your case into our training curriculum to    
Q_77We would like to assure you that we thoroughly investigated your case and found no evidence of wrongdoing by Iraqi force    
Q_78The Iraqi government has agreed to build a public memorial to commemorate the victims of Daesh and the battle for Mosul     
Q_78The Iraqi government has agreed to build three new schools in Mosul to support the city’s recovery from the losses suf      
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

(Dispersion parameter for binomial family taken to be 1)

    Null deviance: 2071.2  on 1503  degrees of freedom
Residual deviance: 1927.9  on 1496  degrees of freedom
AIC: 1943.9

Number of Fisher Scoring iterations: 4

Clustered SEs:

coeftest(q6_5_a_logit, vcov = vcovCL(q6_5_a_logit, cluster = ~ SbjNum))


z test of coefficients:

                                                                                                                              Estimate
(Intercept)                                                                                                                  -0.540512
Q_75We would like to apologize for the death of your loved one.                                                               0.136876
Q_75We would like to express our sympathy for the death of your loved one.                                                    0.041184
Q_76We can offer you a one-time payment of 30 million Iraqi Dinars as compensation for your loss.                             1.255610
Q_77We would like to assure you that we have incorporated the lessons learned from your case into our training curriculum to -0.181020
Q_77We would like to assure you that we thoroughly investigated your case and found no evidence of wrongdoing by Iraqi force  0.032593
Q_78The Iraqi government has agreed to build a public memorial to commemorate the victims of Daesh and the battle for Mosul   0.165501
Q_78The Iraqi government has agreed to build three new schools in Mosul to support the city’s recovery from the losses suf    0.044710
                                                                                                                             Std. Error
(Intercept)                                                                                                                    0.166567
Q_75We would like to apologize for the death of your loved one.                                                                0.127339
Q_75We would like to express our sympathy for the death of your loved one.                                                     0.135198
Q_76We can offer you a one-time payment of 30 million Iraqi Dinars as compensation for your loss.                              0.121705
Q_77We would like to assure you that we have incorporated the lessons learned from your case into our training curriculum to   0.136601
Q_77We would like to assure you that we thoroughly investigated your case and found no evidence of wrongdoing by Iraqi force   0.137889
Q_78The Iraqi government has agreed to build a public memorial to commemorate the victims of Daesh and the battle for Mosul    0.133064
Q_78The Iraqi government has agreed to build three new schools in Mosul to support the city’s recovery from the losses suf     0.134176
                                                                                                                             z value
(Intercept)                                                                                                                  -3.2450
Q_75We would like to apologize for the death of your loved one.                                                               1.0749
Q_75We would like to express our sympathy for the death of your loved one.                                                    0.3046
Q_76We can offer you a one-time payment of 30 million Iraqi Dinars as compensation for your loss.                            10.3169
Q_77We would like to assure you that we have incorporated the lessons learned from your case into our training curriculum to -1.3252
Q_77We would like to assure you that we thoroughly investigated your case and found no evidence of wrongdoing by Iraqi force  0.2364
Q_78The Iraqi government has agreed to build a public memorial to commemorate the victims of Daesh and the battle for Mosul   1.2438
Q_78The Iraqi government has agreed to build three new schools in Mosul to support the city’s recovery from the losses suf    0.3332
                                                                                                                              Pr(>|z|)
(Intercept)                                                                                                                   0.001174
Q_75We would like to apologize for the death of your loved one.                                                               0.282424
Q_75We would like to express our sympathy for the death of your loved one.                                                    0.760654
Q_76We can offer you a one-time payment of 30 million Iraqi Dinars as compensation for your loss.                            < 2.2e-16
Q_77We would like to assure you that we have incorporated the lessons learned from your case into our training curriculum to  0.185115
Q_77We would like to assure you that we thoroughly investigated your case and found no evidence of wrongdoing by Iraqi force  0.813148
Q_78The Iraqi government has agreed to build a public memorial to commemorate the victims of Daesh and the battle for Mosul   0.213587
Q_78The Iraqi government has agreed to build three new schools in Mosul to support the city’s recovery from the losses suf    0.738969
                                                                                                                                
(Intercept)                                                                                                                  ** 
Q_75We would like to apologize for the death of your loved one.                                                                 
Q_75We would like to express our sympathy for the death of your loved one.                                                      
Q_76We can offer you a one-time payment of 30 million Iraqi Dinars as compensation for your loss.                            ***
Q_77We would like to assure you that we have incorporated the lessons learned from your case into our training curriculum to    
Q_77We would like to assure you that we thoroughly investigated your case and found no evidence of wrongdoing by Iraqi force    
Q_78The Iraqi government has agreed to build a public memorial to commemorate the victims of Daesh and the battle for Mosul     
Q_78The Iraqi government has agreed to build three new schools in Mosul to support the city’s recovery from the losses suf      
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

For this binary response, the only significant effect is the 30 million Dinars.