Analysis of Political Leader Evaluations Based on Conflict Concerns
DataManz
Loading the required libraries
if (!require("haven")) install.packages("haven")## Cargando paquete requerido: havenif (!require("dplyr")) install.packages("dplyr")## Cargando paquete requerido: dplyr##
## Adjuntando el paquete: 'dplyr'## The following objects are masked from 'package:stats':
##
## filter, lag## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, unionif (!require("magrittr")) install.packages("magrittr")## Cargando paquete requerido: magrittrif (!require("ggplot2")) install.packages("ggplot2")## Cargando paquete requerido: ggplot2if (!require("factoextra")) install.packages("factoextra")## Cargando paquete requerido: factoextra## Welcome! Want to learn more? See two factoextra-related books at https://goo.gl/ve3WBalibrary(haven)
library(dplyr)
library(magrittr)
library(ggplot2)
library(factoextra)Analysis Overview
This analysis aims to profile participants based on their concerns about international conflicts (variables P2 and P3) and their evaluations of key political leaders (variables VALORALIDERES_1 to VALORALIDERES_4).
The necessary libraries are loaded to facilitate the analysis.
library(haven)
library(dplyr)
library(magrittr)
library(ggplot2)
library(factoextra)The dataset is imported and reviewed to ensure proper loading.
DFWar <- read_sav("C:/Users/Manuel/Desktop/3441.sav")
head(DFWar)Non-response values are converted to missing values to maintain data integrity.
DFWar <- DFWar %>%
mutate(P3 = na_if(P3, 98), P3 = na_if(P3, 99), P4 = na_if(P4, 8), P4 = na_if(P4, 9))
DFWar <- DFWar %>%
mutate(VALORALIDERES_1 = na_if(VALORALIDERES_1, 98), VALORALIDERES_1 = na_if(VALORALIDERES_1, 99),
VALORALIDERES_2 = na_if(VALORALIDERES_2, 98), VALORALIDERES_2 = na_if(VALORALIDERES_2, 99),
VALORALIDERES_3 = na_if(VALORALIDERES_3, 98), VALORALIDERES_3 = na_if(VALORALIDERES_3, 99),
VALORALIDERES_4 = na_if(VALORALIDERES_4, 98), VALORALIDERES_4 = na_if(VALORALIDERES_4, 99))The transformation is verified by counting the missing values, with an expected total of 283.
NA_counts_all <- DFWar %>%
summarize(NA_P3 = sum(is.na(P3)), NA_P4 = sum(is.na(P4)),
NA_VALORALIDERES_1 = sum(is.na(VALORALIDERES_1)),
NA_VALORALIDERES_2 = sum(is.na(VALORALIDERES_2)),
NA_VALORALIDERES_3 = sum(is.na(VALORALIDERES_3)),
NA_VALORALIDERES_4 = sum(is.na(VALORALIDERES_4)))
sum(NA_counts_all)## [1] 283A subset of the relevant columns is created, and missing values are removed.
SubDFWar <- select(DFWar, P2, P3, VALORALIDERES_1:VALORALIDERES_4)
SubDFWarNA <- na.omit(SubDFWar)Hierarchical clustering is performed to identify patterns among participants.
X <- scale(SubDFWarNA, center = FALSE, scale = TRUE)
dj <- dist(X)
cc <- hclust(dj, method = "complete")
plot(cc, main = "Participant Clustering")
The optimal number of clusters is determined using the elbow method and silhouette analysis.
nclusterwar <- fviz_nbclust(SubDFWarNA, kmeans, method = "wss")
nclusterwar
library(cluster)
silhouette_values <- sapply(2:10, function(k) {
kmeans_model <- kmeans(SubDFWarNA, centers = k)
silhouette_avg <- mean(silhouette(kmeans_model$cluster, dist(SubDFWarNA)))
return(silhouette_avg)
})
plot(2:10, silhouette_values, type = "b", pch = 19, frame = FALSE,
xlab = "Number of Clusters",
ylab = "Average Silhouette Coefficient",
main = "Silhouette Method for Optimal Cluster Count")
num_clusters <- 1:5
betweenss_values <- numeric(length(num_clusters))
for (k in num_clusters) {
kmeans_model <- kmeans(SubDFWarNA, k, nstart = 25)
betweenss_values[k] <- kmeans_model$betweenss}
for (k in num_clusters) {
cat("Number of Clusters:", k, "- Between SS:", betweenss_values[k], "\n")}## Number of Clusters: 1 - Between SS: 1.185981e-08
## Number of Clusters: 2 - Between SS: 51712.62
## Number of Clusters: 3 - Between SS: 69066.59
## Number of Clusters: 4 - Between SS: 74733
## Number of Clusters: 5 - Between SS: 79827.43A final cluster analysis is conducted, and the results are summarized.
dist_matrix <- dist(SubDFWarNA)
hc <- hclust(dist_matrix)
cluster_membership <- cutree(hc, k = 3)
SubDFWarNA$Cluster <- cluster_membership
cluster_summary <- aggregate(. ~ cluster_membership, data = SubDFWarNA[, c("P2", "P3", "VALORALIDERES_1", "VALORALIDERES_2", "VALORALIDERES_3", "VALORALIDERES_4")], FUN = mean)
colnames(cluster_summary) <- c("Cluster", "UkraineConflict", "MiddleEastConflict", "PedroSanchez", "AlbertoFeijoo", "YolandaDiaz", "SantiagoAbascal")
print(cluster_summary)## Cluster UkraineConflict MiddleEastConflict PedroSanchez AlbertoFeijoo
## 1 1 2.193660 1.750517 7.192970 2.972433
## 2 2 2.457377 2.432787 1.352459 6.880328
## 3 3 2.397516 2.283456 2.591191 3.896669
## YolandaDiaz SantiagoAbascal
## 1 7.048932 1.343212
## 2 2.032787 6.655738
## 3 2.501976 1.944664Cluster Summaries:
- Cluster 1: Progressive Concerned - High concern for global conflicts, favoring left-wing leaders.
- Cluster 2: Conservative Concerned - Similar concerns but favoring conservative leaders.
- Cluster 3: Generalized Negative Perception - Lower concern levels with distrust in leadership.
Conclusion: The analysis identifies distinct groups based on conflict concerns and political leader evaluations, offering insights into public opinion segmentation.
Analysis of Voting Determinants Based on Household Income, Voting Probability, and Ideological Self-Placement
Introduction This analysis aims to explore the underlying components influencing voting behavior by examining three key variables: household income (INGRESHOG), voting probability (PROBVOTO), and ideological self-placement (ESCIDEOL). The results obtained will help assess the appropriateness and usefulness of the derived solution.
Data Processing The dataset used in this analysis was sourced from a survey file. The following preprocessing steps were conducted:
- Data Selection:
- Only the relevant variables (INGRESHOG, PROBVOTO, ESCIDEOL) were retained for analysis.
library(haven)
DFvote <- read_sav("path/to/datafile.sav")
DFvote <- DFvote %>% select(INGRESHOG, PROBVOTO, ESCIDEOL)- Handling Missing Values:
- Missing values in PROBVOTO and ESCIDEOL were replaced with
NAif they contained values 98 or 99. - Rows with missing values were subsequently omitted.
- Missing values in PROBVOTO and ESCIDEOL were replaced with
DFvote <- DFvote %>%
mutate(PROBVOTO = ifelse(PROBVOTO %in% c(98, 99), NA, PROBVOTO),
ESCIDEOL = ifelse(ESCIDEOL %in% c(98, 99), NA, ESCIDEOL))
DFvote <- na.omit(DFvote)- Data Scaling and Transformation:
- All variables were converted to numeric format to prevent conflicts.
- Standardization was performed to ensure comparability.
DFvote <- DFvote %>%
mutate(INGRESHOG = as.numeric(INGRESHOG),
PROBVOTO = as.numeric(PROBVOTO),
ESCIDEOL = as.numeric(ESCIDEOL))
DFvote_scaled <- DFvote %>%
mutate(INGRESHOG = scale(INGRESHOG),
PROBVOTO = scale(PROBVOTO),
ESCIDEOL = scale(ESCIDEOL))Exploratory Analysis An initial frequency analysis of the variables revealed a high probability of certain individuals voting, which could be an important factor in subsequent analysis.
variables <- c("INGRESHOG", "PROBVOTO", "ESCIDEOL")
for (variable in variables) {
cat("Frequency of repeated values for", variable, ":\n")
print(table(DFvote[[variable]]))
cat("\n")
}Correlation Analysis A correlation heatmap was generated to examine the relationships between variables. The results indicated minimal correlations, suggesting potential data simplicity and questioning the suitability of factorial analysis.
library(ggplot2)
library(reshape2)
matcor <- cor(DFvote_scaled)
matriz_cor_melt <- melt(matcor)
colnames(matriz_cor_melt) <- c("variable_x", "variable_y", "correlation")
ggplot(data = matriz_cor_melt, aes(x = variable_x, y = variable_y, fill = correlation)) +
geom_tile() +
geom_text(aes(label = round(correlation, 2)), color = "black") +
scale_fill_gradient2(low = "blue", high = "red", mid = "white", midpoint = 0) +
theme_minimal() +
theme(axis.text.x = element_text(angle = 45, vjust = 1)) +
coord_fixed()Factor Analysis Given the results, a factorial analysis was conducted cautiously. The number of factors was determined using the eigenvalue criterion, which suggested that two factors should be retained.
library(psych)
pca_result <- prcomp(matcor, scale = TRUE)
eigenvalues <- pca_result$sdev^2
factores_significativos <- sum(eigenvalues > 1)- Factor Extraction:
- A two-factor model was estimated using the minimum residuals method with varimax rotation.
modelovote <- fa(DFvote_scaled, nfactors = 2, rotate = "varimax", fm = "minres")
print(modelovote$loadings, cut = 0)Conclusion The factorial analysis provides valuable insights into the determinants of voting behavior:
- Factor 1: Socioeconomic and Electoral Participation
Dimension
- Significant relationship between household income and voting probability.
- Suggests lower-income individuals may be more inclined to participate in voting to influence policies affecting their economic situation.
- Factor 2: Ideological Dimension
- Highlights the influence of ideological self-placement on electoral decisions, albeit with a lower explanatory power.
Final Remarks This analysis underscores the importance of understanding both socioeconomic and ideological dimensions in electoral behavior. Although the data quality poses some limitations, the extracted factors offer meaningful insights for policymakers and political strategists.
3
Modeling the variable P1 using age, gender, and ideological self-placement. Analysis of Voting Determinants Based on Household Income, Voting Probability, and Ideological Self-Placement
Introduction This analysis aims to explore the underlying components influencing voting behavior by examining three key variables: household income (INGRESHOG), voting probability (PROBVOTO), and ideological self-placement (ESCIDEOL). The results obtained will help assess the appropriateness and usefulness of the derived solution.
Data Processing The dataset used in this analysis was sourced from a survey file. The following preprocessing steps were conducted:
- Data Selection:
- Only the relevant variables (INGRESHOG, PROBVOTO, ESCIDEOL) were retained for analysis.
library(haven)
DFvote <- read_sav("path/to/datafile.sav")
DFvote <- DFvote %>% select(INGRESHOG, PROBVOTO, ESCIDEOL)- Handling Missing Values:
- Missing values in PROBVOTO and ESCIDEOL were replaced with
NAif they contained values 98 or 99. - Rows with missing values were subsequently omitted.
- Missing values in PROBVOTO and ESCIDEOL were replaced with
DFvote <- DFvote %>%
mutate(PROBVOTO = ifelse(PROBVOTO %in% c(98, 99), NA, PROBVOTO),
ESCIDEOL = ifelse(ESCIDEOL %in% c(98, 99), NA, ESCIDEOL))
DFvote <- na.omit(DFvote)- Data Scaling and Transformation:
- All variables were converted to numeric format to prevent conflicts.
- Standardization was performed to ensure comparability.
DFvote <- DFvote %>%
mutate(INGRESHOG = as.numeric(INGRESHOG),
PROBVOTO = as.numeric(PROBVOTO),
ESCIDEOL = as.numeric(ESCIDEOL))
DFvote_scaled <- DFvote %>%
mutate(INGRESHOG = scale(INGRESHOG),
PROBVOTO = scale(PROBVOTO),
ESCIDEOL = scale(ESCIDEOL))Exploratory Analysis An initial frequency analysis of the variables revealed a high probability of certain individuals voting, which could be an important factor in subsequent analysis.
variables <- c("INGRESHOG", "PROBVOTO", "ESCIDEOL")
for (variable in variables) {
cat("Frequency of repeated values for", variable, ":\n")
print(table(DFvote[[variable]]))
cat("\n")
}Correlation Analysis A correlation heatmap was generated to examine the relationships between variables. The results indicated minimal correlations, suggesting potential data simplicity and questioning the suitability of factorial analysis.
library(ggplot2)
library(reshape2)
matcor <- cor(DFvote_scaled)
matriz_cor_melt <- melt(matcor)
colnames(matriz_cor_melt) <- c("variable_x", "variable_y", "correlation")
ggplot(data = matriz_cor_melt, aes(x = variable_x, y = variable_y, fill = correlation)) +
geom_tile() +
geom_text(aes(label = round(correlation, 2)), color = "black") +
scale_fill_gradient2(low = "blue", high = "red", mid = "white", midpoint = 0) +
theme_minimal() +
theme(axis.text.x = element_text(angle = 45, vjust = 1)) +
coord_fixed()Factor Analysis Given the results, a factorial analysis was conducted cautiously. The number of factors was determined using the eigenvalue criterion, which suggested that two factors should be retained.
library(psych)
pca_result <- prcomp(matcor, scale = TRUE)
eigenvalues <- pca_result$sdev^2
factores_significativos <- sum(eigenvalues > 1)- Factor Extraction:
- A two-factor model was estimated using the minimum residuals method with varimax rotation.
modelovote <- fa(DFvote_scaled, nfactors = 2, rotate = "varimax", fm = "minres")
print(modelovote$loadings, cut = 0)Conclusion The factorial analysis provides valuable insights into the determinants of voting behavior:
- Factor 1: Socioeconomic and Electoral Participation
Dimension
- Significant relationship between household income and voting probability.
- Suggests lower-income individuals may be more inclined to participate in voting to influence policies affecting their economic situation.
- Factor 2: Ideological Dimension
- Highlights the influence of ideological self-placement on electoral decisions, albeit with a lower explanatory power.
Logistic Regression Model Analysis
Data Preparation and Model Fitting
subdfsanchez <- read_sav("path/to/datafile.sav") %>%
select(P1, EDAD, SEXO, ESCIDEOL) %>%
mutate(ESCIDEOL = replace(ESCIDEOL, ESCIDEOL %in% c(98, 99), NA)) %>%
na.omit()
subdfsanchez$P1 <- ifelse(subdfsanchez$P1 == 2, 0, subdfsanchez$P1)
modeloglmsanchez <- glm(P1 ~ EDAD + SEXO + ESCIDEOL, data = subdfsanchez, family = "binomial")
summary(modeloglmsanchez)Evaluation of Model Fit
hoslem_test <- hoslem.test(subdfsanchez$P1, fitted(modeloglmsanchez), g = 5)
hoslem_testPredictor Importance
- Gender (SEXO): Most significant predictor with a negative impact.
- Age (EDAD): Positive association with voting probability.
- Ideological Self-Placement (ESCIDEOL): Lesser significance in the model.
Final Remarks This analysis underscores the importance of understanding both socioeconomic and ideological dimensions in electoral behavior. Although the data quality poses some limitations, the extracted factors and regression models offer meaningful insights for policymakers and political strategists.
4
Analysis of Voting Determinants Based on Household Income, Voting Probability, and Ideological Self-Placement
Introduction This analysis aims to explore the underlying components influencing voting behavior by examining three key variables: household income (INGRESHOG), voting probability (PROBVOTO), and ideological self-placement (ESCIDEOL). The results obtained will help assess the appropriateness and usefulness of the derived solution.
Data Processing The dataset used in this analysis was sourced from a survey file. The following preprocessing steps were conducted:
- Data Selection:
- Only the relevant variables (INGRESHOG, PROBVOTO, ESCIDEOL) were retained for analysis.
library(haven)
DFvote <- read_sav("path/to/datafile.sav")
DFvote <- DFvote %>% select(INGRESHOG, PROBVOTO, ESCIDEOL)- Handling Missing Values:
- Missing values in PROBVOTO and ESCIDEOL were replaced with
NAif they contained values 98 or 99. - Rows with missing values were subsequently omitted.
- Missing values in PROBVOTO and ESCIDEOL were replaced with
DFvote <- DFvote %>%
mutate(PROBVOTO = ifelse(PROBVOTO %in% c(98, 99), NA, PROBVOTO),
ESCIDEOL = ifelse(ESCIDEOL %in% c(98, 99), NA, ESCIDEOL))
DFvote <- na.omit(DFvote)- Data Scaling and Transformation:
- All variables were converted to numeric format to prevent conflicts.
- Standardization was performed to ensure comparability.
DFvote <- DFvote %>%
mutate(INGRESHOG = as.numeric(INGRESHOG),
PROBVOTO = as.numeric(PROBVOTO),
ESCIDEOL = as.numeric(ESCIDEOL))
DFvote_scaled <- DFvote %>%
mutate(INGRESHOG = scale(INGRESHOG),
PROBVOTO = scale(PROBVOTO),
ESCIDEOL = scale(ESCIDEOL))Exploratory Analysis An initial frequency analysis of the variables revealed a high probability of certain individuals voting, which could be an important factor in subsequent analysis.
variables <- c("INGRESHOG", "PROBVOTO", "ESCIDEOL")
for (variable in variables) {
cat("Frequency of repeated values for", variable, ":\n")
print(table(DFvote[[variable]]))
cat("\n")
}Regression Model to Explain Survey Duration
Data Preparation and Model Fitting
DFTIME <- read_sav("path/to/datafile.sav")
subftime <- DFTIME %>% select(EDAD, SEXO, ESCIDEOL, TIPO_TEL, IA_E3)
subftime$ESCIDEOL[subftime$ESCIDEOL %in% c(98, 99)] <- NA
subftime <- na.omit(subftime)
subftime$SEXO <- ifelse(subftime$SEXO == 2, 0, 1)
subftime$TIPO_TEL <- ifelse(subftime$TIPO_TEL == 2, 1, 0)
subftime$EDAD <- factor(subftime$EDAD)
subftime$SEXO <- factor(subftime$SEXO)
subftime$ESCIDEOL <- factor(subftime$ESCIDEOL)
subftime$TIPO_TEL <- factor(subftime$TIPO_TEL)ANOVA Analysis
modeloanova <- aov(IA_E3 ~ EDAD + SEXO + ESCIDEOL + TIPO_TEL, data = subftime)
summary(modeloanova)The variable EDAD is the only one that shows a significant effect on the duration of the survey, while the other variables (SEXO, ESCIDEOL, TIPO_TEL) do not have a significant effect according to the ANOVA analysis. This indicates that the duration of the survey is primarily influenced by the respondents’ age.
Linear Regression Analysis
modelo <- lm(IA_E3 ~ EDAD + SEXO + ESCIDEOL + TIPO_TEL, data = subftime)
summary(modelo)The results suggest that age, ideological self-placement, and telephone type are significant factors influencing survey duration, whereas gender appears to have minimal and non-significant influence. Removing gender, we obtain a refined model:
modelo2 <- lm(IA_E3 ~ EDAD + ESCIDEOL + TIPO_TEL, data = subftime)
summary(modelo2)Conclusion
- Age (EDAD) has the most significant effect on survey duration, with older individuals taking longer to complete the survey.
- Ideological self-placement (ESCIDEOL) also plays a role, with those identifying further along the ideological spectrum taking slightly longer.
- Telephone type (TIPO_TEL), while not significant in ANOVA, showed a significant coefficient in the regression model, suggesting a potential influence not captured in ANOVA.
- Gender (SEXO) does not significantly influence survey duration.
Model Utility
- Resource Optimization: Helps identify potential over- or under-utilization of resources such as personnel and materials.
- Training and Development: Insights from the model can inform personalized training programs for interviewers.
- Quality Evaluation: Comparing predicted vs. actual survey durations can provide insights into data quality.
- Strategic Planning: Facilitates planning by predicting expected survey completion times under different conditions.
- Trend Identification: Analyzing variations in survey duration can reveal patterns that aid in operational efficiency.
- Performance Comparison: Enables comparison of interviewer performance across regions to establish best practices.