Final Project: Exploring Factors Associated with Divorce: Agreement on Values and Awareness of Spouse’s Stress
Bella Sly
18 May, 2025
Introduction
Provide a brief introduction to your research question. Summarize the empirical articles you reviewed.
Divorce is something many couples go through, and researchers have been trying to figure out what makes some marriages last while others end. Some things that can affect a marriage include how well couples understand each other and whether they share similar values.
In this project, I’m looking at whether two things can help predict if a couple will get divorced:
If they have similar values about personal freedom (Q12), and If one partner knows what the other is currently stressed about (Q27). These two things might show how close a couple is and how well they work together. If couples understand each other and agree on important things, they might be less likely to get divorced.
Literature Review
Article 1 Summary
Summarize the empirical article you selected. Discuss the key findings. Cleek, M. G., & Pearson, T. A. (1985) This study explored the reasons divorced individuals believe their marriages ended. The researchers surveyed 86 recently divorced people and asked them to identify the causes of their divorce. The most frequently cited reasons included lack of communication, incompatibility, infidelity, and growing apart. One key takeaway was that many of these causes were linked together rather than isolated—for example, poor communication often went hand in hand with emotional distance or mismatched values. The study supports the idea that successful marriages depend on shared values and open communication, both of which relate to the variables used in this project: similarity in personal values and awareness of a spouse’s stress.
Article 2 Summary
Summarize the empirical article you selected. Discuss the key findings. Lowenstein, L. F. (2005) This article reviews a range of research on the causes and characteristics of divorce, identifying the most common patterns seen in recent studies. Lowenstein highlights emotional detachment, unresolved conflicts, lack of empathy, and differences in life goals or values as major contributors to marital breakdown. He emphasizes that many divorces are caused by partners failing to stay emotionally connected or communicate effectively during stressful periods. These findings align closely with the current study’s focus on whether couples share values around personal freedom and understand each other’s sources of stress—both of which are strong indicators of emotional engagement and compatibility in a relationship.
Hypothesis
State your directional hypothesis. Specify the expected relationship between your variables. This section should be clear and concise. You need one hypothesis for IV1, one hypothesis for IV2, and one hypothesis for a predicted interaction. If you are running a logistic regression, you do not need a hypothesis for an interaction.
Hypothesis 1 - I predict that people who agree more with the statement “My wife and I have similar values in terms of personal freedom” will be less likely to get divorced. Hypothesis 2 - I predict that people who agree more with the statement “I know what my wife’s current sources of stress are” will also be less likely to get divorced. # Method
Sample
Describe the sample used in your study. Include details about the population, sample size, and any relevant demographic information.
The dataset comes from the “Split or Stay Divorce Predictor” on Kaggle. It includes responses from 170 participants who answered questions related to their marriage. Each person rated how much they agreed or disagreed with 54 statements on a scal from 0 (Strongly Disagree) to 4 (Strongly Agree). These questions cover different parts of a relationship, such as communication, values, emotional connection, and conflict. The main outcome in the dataset is whether the person is divorced (1) or not divorced (0). While the dataset doesn’t include detailed demographic information like age or gender, it gives insight into how relationship beliefs and behaviors may relate to the chance of divorce.
Variables and Operationalization
List your independent and dependent variables. Explain how each variable was operationalized, including the range for continuous variables and levels for categorical variables.
IV: Q2 “My wife and I have similar values in terms of personal freedom” This is a continuous variable measured on a scale from 0-4, where 0 means Strongly Disagree and 4 means Strongly Agree. Higher scores indicate greater agreement that the couple shares similar values about personal freedom. IV : Q27: “I know what my wife’s current sources of stress are” This is a contiuous variable on the same 0-4 scale. Higher scores show that the person feels more aware of their spouses current stressors.” ## Loading Required Libraries
# Load your dataset in this chunk
library(readr)
divorce_data<- read_delim("/Users/BellaSly/Downloads/divorce/divorce.csv", delim = ";")
colnames(divorce_data)## [1] "Atr1" "Atr2" "Atr3" "Atr4" "Atr5" "Atr6" "Atr7" "Atr8" "Atr9"
## [10] "Atr10" "Atr11" "Atr12" "Atr13" "Atr14" "Atr15" "Atr16" "Atr17" "Atr18"
## [19] "Atr19" "Atr20" "Atr21" "Atr22" "Atr23" "Atr24" "Atr25" "Atr26" "Atr27"
## [28] "Atr28" "Atr29" "Atr30" "Atr31" "Atr32" "Atr33" "Atr34" "Atr35" "Atr36"
## [37] "Atr37" "Atr38" "Atr39" "Atr40" "Atr41" "Atr42" "Atr43" "Atr44" "Atr45"
## [46] "Atr46" "Atr47" "Atr48" "Atr49" "Atr50" "Atr51" "Atr52" "Atr53" "Atr54"
## [55] "Class"Descriptive Statistics
Present the descriptive statistics for your variables. Include appropriate measures of central tendency (mean, median), variability (standard deviation, range), and frequency distributions where applicable. Use R code chunks to generate and display your results.
## vars n mean sd median trimmed mad min max range skew kurtosis se
## Atr12 1 170 1.65 1.47 1.5 1.57 2.22 0 4 4 0.20 -1.47 0.11
## Atr27 2 170 1.40 1.46 1.0 1.29 1.48 0 4 4 0.36 -1.50 0.11## < table of extent 0 >## [1] "Atr1" "Atr2" "Atr3" "Atr4" "Atr5" "Atr6" "Atr7" "Atr8" "Atr9"
## [10] "Atr10" "Atr11" "Atr12" "Atr13" "Atr14" "Atr15" "Atr16" "Atr17" "Atr18"
## [19] "Atr19" "Atr20" "Atr21" "Atr22" "Atr23" "Atr24" "Atr25" "Atr26" "Atr27"
## [28] "Atr28" "Atr29" "Atr30" "Atr31" "Atr32" "Atr33" "Atr34" "Atr35" "Atr36"
## [37] "Atr37" "Atr38" "Atr39" "Atr40" "Atr41" "Atr42" "Atr43" "Atr44" "Atr45"
## [46] "Atr46" "Atr47" "Atr48" "Atr49" "Atr50" "Atr51" "Atr52" "Atr53" "Atr54"
## [55] "Class"Statistical Analysis
Analysis
Perform your chosen analysis. Make sure your output shows.
logistic_model <- glm(Class ~ Atr12 + Atr27, data = divorce_data, family = binomial)
summary(logistic_model)##
## Call:
## glm(formula = Class ~ Atr12 + Atr27, family = binomial, data = divorce_data)
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -4.8327 0.8823 -5.478 4.31e-08 ***
## Atr12 1.9211 0.5344 3.595 0.000325 ***
## Atr27 2.0008 0.6595 3.034 0.002415 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 235.647 on 169 degrees of freedom
## Residual deviance: 36.477 on 167 degrees of freedom
## AIC: 42.477
##
## Number of Fisher Scoring iterations: 8Post-hoc Power Analysis
Run a post-hoc power analysis with the pwr package. Use
the pwr.f2.test function for multiple regression power
analysis.
library(pwr)
# Calculate effect size f2 from pseudo R²
pseudo_r2 <- 1 - (36.477 / 235.647)
f2 <- pseudo_r2 / (1 - pseudo_r2)
#Post-hoc power analysis
pwr.f2.test(
u = 2, # number of predictors
v = 170 - 2 - 1, # degrees of freedom = N - predictors - 1
f2 = f2, # effect size
sig.level = 0.05 # alpha level
)##
## Multiple regression power calculation
##
## u = 2
## v = 167
## f2 = 5.460153
## sig.level = 0.05
## power = 1Results Interpretation
Results are interpreted clearly using APA style; connection to hypothesis is made; statistical significance and practical implications are addressed; power level is addressed.
A logistic regression was conducted to examine whether agreement with two relationship statements predicted the likelihood of being divorced. The model was statistically significant, χ²(2) = 199.17, p < .001, indicating that the predictors reliably distinguished between those who were divorced and not divorced.
Both predictors were significant:
Atr12, which asked if the participant and their spouse have similar values around personal freedom, had a significant positive effect on divorce likelihood, B = 1.92, SE = 0.53, p < .001. This means that higher agreement with this statement was associated with a higher chance of being divorced. Atr27, which asked if the participant knows what their spouse’s current sources of stress are, also had a significant positive effect, B = 2.00, SE = 0.66, p = .002. This suggests that knowing a spouse’s stress was also linked to a higher chance of divorce. The post-hoc power analysis showed a very large effect size (f² = 5.46), with a power of 1.00. This means the study had more than enough power to detect an effect, so the results are likely to be reliable.
Although these results might seem surprising, they suggest that agreement on values like freedom and being very in tune with a partner’s stress might not always protect against divorce. This could be due to over-involvement, mismatched expectations, or deeper underlying issues not captured by the questions.
Graph and Table
Include at least one table and one graph that effectively summarize your analysis and findings. Use R code chunks to generate these visualizations.
library(ggplot2)
library(knitr)
# Make sure Class is numeric 0/1 for logistic regression
divorce_data$Class <- as.numeric(as.character(divorce_data$Class))
# Fit logistic regression model with Class as dependent variable
logistic_model <- glm(Class ~ Atr12, data = divorce_data, family = binomial)
# Add predicted probabilities to data frame
divorce_data$predicted_prob <- predict(logistic_model, type = "response")
# APA-style ggplot theme
theme_apa <- function() {
theme_minimal(base_size = 12) +
theme(
text = element_text(family = "serif", color = "black"),
axis.title = element_text(size = 14, face = "bold"),
axis.text = element_text(size = 12),
panel.grid.major = element_blank(),
panel.grid.minor = element_blank(),
axis.line = element_line(color = "black"),
axis.ticks = element_line(color = "black"),
legend.position = "top",
legend.title = element_text(size = 12, face = "bold"),
legend.text = element_text(size = 12),
plot.title = element_text(size = 16, face = "bold", hjust = 0.5)
)
}
# Create the plot
p <- ggplot(divorce_data, aes(x = Atr12, y = predicted_prob)) +
geom_point(alpha = 0.5) +
geom_smooth(method = "loess") +
labs(
title = "Predicted Probability of Divorce by Agreement with Similar Values (Atr12)",
x = "Agreement with Similar Values (0 to 4)",
y = "Predicted Probability of Divorce"
) +
theme_apa()
print(p)## `geom_smooth()` using formula = 'y ~ x'
# Display table of first 6 rows of Atr12 and predicted probabilities
kable(head(divorce_data[, c("Atr12", "predicted_prob")]), caption = "Predicted probabilities by Atr12")| Atr12 | predicted_prob |
|---|---|
| 0 | 0.0067717 |
| 3 | 0.9887549 |
| 4 | 0.9995153 |
| 3 | 0.9887549 |
| 1 | 0.1378456 |
| 2 | 0.7894492 |
library(dplyr)
library(ggplot2)
# Correct median split for Atr12
divorce_data <- divorce_data %>%
mutate(Atr12_split = ifelse(Atr12 > median(Atr12), "High Agreement", "Low Agreement"))
# Add predicted probabilities from logistic model
divorce_data$predicted_prob <- predict(logistic_model, type = "response")
# Plot the interaction using the median split
ggplot(divorce_data, aes(x = Atr27, y = predicted_prob, color = Atr12_split)) +
geom_point(alpha = 0.6) +
geom_smooth(method = "loess", se = FALSE) +
labs(
title = "Predicted Probability of Divorce by Knowing Spouse’s Stress (Atr27)",
x = "Agreement with Knowing Spouse’s Stress (0-4)",
y = "Predicted Probability of Divorce",
color = "Agreement on Similar Values (Atr12)"
) +
scale_color_manual(values = c("High Agreement" = "green", "Low Agreement" = "orange")) +
theme_apa()
# Example R code for creating a table
# Create a summary table by Species
library(dplyr)
# Create a median split for Atr12
divorce_data <- divorce_data %>%
mutate(Atr12_split = ifelse(Atr12 > median(Atr12), "High Agreement", "Low Agreement"))
# Create summary table by Atr12_split
summary_table <- divorce_data %>%
group_by(Atr12_split) %>%
summarise(
PredictedProb_Mean = mean(predicted_prob, na.rm = TRUE),
PredictedProb_SD = sd(predicted_prob, na.rm = TRUE),
PredictedProb_Min = min(predicted_prob, na.rm = TRUE),
PredictedProb_Max = max(predicted_prob, na.rm = TRUE)
)
# View the summary table
print(summary_table)## # A tibble: 2 × 5
## Atr12_split PredictedProb_Mean PredictedProb_SD PredictedProb_Min
## <chr> <dbl> <dbl> <dbl>
## 1 High Agreement 0.940 0.0895 0.789
## 2 Low Agreement 0.0484 0.0614 0.00677
## # ℹ 1 more variable: PredictedProb_Max <dbl>##
## Attaching package: 'kableExtra'## The following object is masked from 'package:dplyr':
##
## group_rowskable(summary_table, caption = "Descriptive Statistics of Predicted Divorce Probability by Agreement with Similar Values (Atr12)") %>%
kable_styling(bootstrap_options = c("striped", "hover", "condensed"), full_width = FALSE)| Atr12_split | PredictedProb_Mean | PredictedProb_SD | PredictedProb_Min | PredictedProb_Max |
|---|---|---|---|---|
| High Agreement | 0.9398284 | 0.0894998 | 0.7894492 | 0.9995153 |
| Low Agreement | 0.0484069 | 0.0613851 | 0.0067717 | 0.1378456 |
Discussion
Discuss the implications of your results for psychological theory or practice. Address the following points:
- Implications: What do your findings mean in the context of existing research?
These results suggest that merely sharing values or being aware of a partner’s stress does not guarantee marital stability. Instead, the quality of how partners manage and respond to these shared values and stressors may be critical. This has important implications for psychological practice, particularly marriage counseling, where interventions might focus more on coping strategies, communication, and stress management rather than only on fostering shared values or awareness.
- Limitations: Identify any limitations of your study. How might these limitations have affected your results?
This study has several limitations. First, the cross-sectional design limits causal inference; we cannot determine whether agreement with these statements leads to divorce or if divorced individuals retrospectively interpret these dimensions differently. Second, the sample size was modest, and the data were self-reported, introducing potential biases. Third, the operationalization of the variables using single-item measures might not fully capture the complexity of marital dynamics.
- Future Directions: Suggest potential future research directions based on your findings.
Longitudinal research is needed to clarify causality. Future studies should include more detailed measures of communication and conflict resolution to better understand what drives divorce.
References
List the articles you reviewed in APA format. Do not worry about the indentations.
Cleek, M. G., & Pearson, T. A. (1985). Perceived causes of divorce: An analysis of interrelationships. Journal of Marriage and the Family, 47(1), 179–183. Lowenstein, L. F. (2005). Causes and associated features of divorce as seen by recent research. Journal of Divorce & Remarriage, 42(3–4), 87–102.