Final Project - Student Depression Dataset
PSYCH-214-5291
Allison Cavanagh
23 May, 2025
Introduction
This study investigates whether academic performance predicts depression status among students, and further examines if sleep duration modifies this relationship. This is an especially important subject because depression is a significant mental health concern across the world that can negatively impact a student’s academic success and overall well being. This research aims to identify the relevant factors that influence depression risk to students.
Research question:
Does academic performance predict depression status among students? Does sleep duration interact with this relationship and change it?
Empirical article review:
Berger et al. (2019) examined the role of sleep in moderating the relationship between home chaos and academic achievement in young children. The study included 103 kindergartners and first graders, ranging from 5 to 7 years old. The sample consisted of 51% girls and 49% boys. The participants were recruited from a larger study, and demographic information was also collected. Home chaos was assessed through parental reports. Sleep duration was measured using actigraphs worn by children for 5 consecutive nights of school. Academic achievement was measured using the Applied Problems and Passage Comprehension tests from the Woodcock-Johnson III Tests of Achievement. The researchers hypothesized that children from low-chaos homes and with longer sleep duration/ higher sleep efficiency would show higher academic achievement. Further, they predicted negative effects of home chaos on academic achievement would be the strongest for children with less sleep/ lower sleep efficiency. The findings revealed that longer sleep durations helped decrease the negative impact of home chaos on academic achievement, where children who had more sleep performed better despite high levels of home chaos. This study highlights the role of sleep on academic outcomes, which is relevant to my research question in examining how sleep duration might influence the relationship between academic performance and depression in students.
Verboom et al. (2015) conducted a longitudinal study to examine the relationship between depressive problems and three types of functioning in adolescents: academic performance, social well-being, and social problems. The study involved 2,230 participants aged 10-18. It used questionnaires and path analyses to investigate these dynamics. The researchers hypothesize that depressive problems would influence social functioning and academic performance over time. Additionally, they expected a one-way relationship where poor academic performance would contribute to depressive problems, but depressive problems would not strongly predict academic outcomes. The study found bidirectional relationships between depressive problems and academic performance, but these relationships were gender-specific. For girls, depressive problems and academic performance influenced each other over time. Additionally, depressive symptoms were associated with poorer academic performance and vice versa. However, with boys, the relationship between depressive problems and academic performance was not bidirectional; depressive symptoms did not significantly predict changes in academic performance, nor did academic performance predict depressive symptoms over time. Their study generally concludes that the link between depression and academic outcome grants different results for boys and girls, with girls being more vulnerable to the impact of depression on their academic success. These findings provide important context for investigating how academic performance may predict depression status in students. It also suggests that gender differences may be an important factor in this relationship. However, it supports my research hypothesis in the scope that depression and academic performance are closely linked.
Hypothesis
It is hypothesized that lower academic performance (CGPA) will be associated with a higher likelihood of depression and that this relationship will be stronger for students who report shorter sleep duration.
Method
Sample:
The dataset contains 27,901 participants aged 18-59 years old, with about 44% female and about 56% male. The dataset included responses from participants across 52 regions and 14 different professions.
Variables:
The independent variables in this study include academic performance and sleep duration. Academic performance is operationalized as the Cumulative Grade Point Average (CGPA), which ranges from 0.0 to 4.0, with higher scores indicating better academic achievement. Sleep duration is measured as the average number of hours of sleep per day, with an expected range of 0 to 12 hours. The dependent variable is depression status, which is operationalized through self-report and categorized as either “Yes” or “No” to indicate whether a student is experiencing depression.
Descriptive Statistics
library(dplyr)##
## Attaching package: 'dplyr'## The following objects are masked from 'package:stats':
##
## filter, lag## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, unionstudent_data <- read.csv("/Users/allisoncavanagh/Downloads/Student Depression Dataset.csv", stringsAsFactors = FALSE)
student_data <- student_data %>%
mutate(SleepHours = case_when(
Sleep.Duration == "less than 5 hours" ~ 4,
Sleep.Duration == "5-6 hours" ~ 5.5,
Sleep.Duration == "7-8 hours" ~ 7.5,
Sleep.Duration == "more than 8 hours" ~ 9,
TRUE ~ NA_real_
))
clean_data <- student_data %>%
filter(!is.na(CGPA), !is.na(SleepHours), !is.na(Depression))
summary_stats <- clean_data %>%
summarise(
N = n(),
CGPA_mean = mean(CGPA),
CGPA_median = median(CGPA),
CGPA_sd = sd(CGPA),
CGPA_min = min(CGPA),
CGPA_max = max(CGPA),
Sleep_mean = mean(SleepHours),
Sleep_median = median(SleepHours),
Sleep_sd = sd(SleepHours),
Sleep_min = min(SleepHours),
Sleep_max = max(SleepHours)
)
print(summary_stats)## N CGPA_mean CGPA_median CGPA_sd CGPA_min CGPA_max Sleep_mean
## 1 13529 7.68732 7.83 1.467534 0 10 6.585963
## Sleep_median Sleep_sd Sleep_min Sleep_max
## 1 7.5 0.9963351 5.5 7.5depression_freq <- table(clean_data$Depression)
depression_pct <- prop.table(depression_freq) * 100
print(depression_freq)##
## 0 1
## 5641 7888print(round(depression_pct, 1))##
## 0 1
## 41.7 58.3table(student_data$DepressionStatus)## < table of extent 0 >student_data <- read.csv("/Users/allisoncavanagh/Downloads/Student Depression Dataset.csv", stringsAsFactors = FALSE)**The dataset included 13,529 participants. The academic performance variable, operationalized as CGPA, had a mean of 7.69 (SD = 1.47), with a median of 7.83. The CGPA scores ranged from 0 to 10, indicating a wide range of academic outcomes. Sleep duration averaged 6.59 hours per day (SD = 1.00), with a median of 7.5 hours. Reported sleep hours ranged from 5.5 to 7.5 hours among the participants. Regarding mental health, 7,888 participants (58.3%) reported experiencing symptoms of depression, while 5,641 participants (41.7%) did not.
Statistical Analysis
model <- glm(Depression ~ CGPA * Sleep.Duration, data = student_data, family = binomial)
summary(model)##
## Call:
## glm(formula = Depression ~ CGPA * Sleep.Duration, family = binomial,
## data = student_data)
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -0.034163 0.138569 -0.247 0.80527
## CGPA 0.040504 0.017736 2.284 0.02239 *
## Sleep.Duration7-8 hours 0.168811 0.186752 0.904 0.36603
## Sleep.DurationLess than 5 hours 0.489342 0.184408 2.654 0.00796 **
## Sleep.DurationMore than 8 hours -0.155549 0.192931 -0.806 0.42010
## Sleep.DurationOthers -3.334712 2.602606 -1.281 0.20009
## CGPA:Sleep.Duration7-8 hours -0.007936 0.023901 -0.332 0.73985
## CGPA:Sleep.DurationLess than 5 hours -0.021839 0.023663 -0.923 0.35605
## CGPA:Sleep.DurationMore than 8 hours -0.010694 0.024791 -0.431 0.66621
## CGPA:Sleep.DurationOthers 0.403170 0.336628 1.198 0.23105
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 37859 on 27900 degrees of freedom
## Residual deviance: 37567 on 27891 degrees of freedom
## AIC: 37587
##
## Number of Fisher Scoring iterations: 4table(student_data$Sleep.Duration)##
## 5-6 hours 7-8 hours Less than 5 hours More than 8 hours
## 6183 7346 8310 6044
## Others
## 18The intercept (p = 0.805) is not significant. The CGPA is positive and significant with an estimate of 0.0405 and p = 0.022. This means that more each unit increase in CGPA, the log odds of depression will increase slightly. The coefficient for “Less than 5 hours” of sleep is 0.4893 and is statistically significant with p = 0.008. This means that compared to students who sleep 5-6 hours, those who sleep less than 5 hours have higher odds of reporting depression. Other sleep categories are not significant, meaning they do not significantly moderate the realtionship between CGPA and depression, within this dataset.
Post-hoc Power Analysis
library(pwr)
h <- 0.234628 # Effect size (Cohen's h)
n <- 27901 # Total sample size
sig_level <- 0.05 # Significance level (alpha)
power_result <- pwr.2p.test(h = h, n = n/2, sig.level = sig_level, alternative = "two.sided")
print(power_result)##
## Difference of proportion power calculation for binomial distribution (arcsine transformation)
##
## h = 0.234628
## n = 13950.5
## sig.level = 0.05
## power = 1
## alternative = two.sided
##
## NOTE: same sample sizesThe effect size for the difference in depression proportions between groups were found to be 0.235, a small to medium effect size. Using 27,901 participants, a significance level of 0.05, and a two sided test, the post-hoc power analysis showed the statistical power to be 1.00 or 100%. This means that the likelihood of a Type II error is very low, meaning the findings frm my logistic regression analysis are statisically reliable.
Tables and Graphs
library(dplyr)
student_data <- read.csv("/Users/allisoncavanagh/Downloads/Student Depression Dataset.csv", stringsAsFactors = FALSE)
model <- glm(Depression ~ CGPA * Sleep.Duration, data = student_data, family = binomial)
coef_df <- summary(model)$coefficients
selected_rows <- grep("Sleep.Duration|CGPA", rownames(coef_df))
selected_coefs <- coef_df[selected_rows, ]
clean_names <- rownames(selected_coefs)
clean_names <- gsub("Sleep.Duration", "", clean_names)
clean_names <- gsub("CGPA:", "CGPA x ", clean_names)
clean_names <- gsub(":", " x ", clean_names)
summary_table <- data.frame(
Term = clean_names,
Estimate = round(selected_coefs[, "Estimate"], 3),
Odds_Ratio = round(exp(selected_coefs[, "Estimate"]), 3),
Std_Error = round(selected_coefs[, "Std. Error"], 3),
p_value = round(selected_coefs[, "Pr(>|z|)"], 4),
stringsAsFactors = FALSE
)
print(summary_table)## Term Estimate
## CGPA CGPA 0.041
## Sleep.Duration7-8 hours 7-8 hours 0.169
## Sleep.DurationLess than 5 hours Less than 5 hours 0.489
## Sleep.DurationMore than 8 hours More than 8 hours -0.156
## Sleep.DurationOthers Others -3.335
## CGPA:Sleep.Duration7-8 hours CGPA x 7-8 hours -0.008
## CGPA:Sleep.DurationLess than 5 hours CGPA x Less than 5 hours -0.022
## CGPA:Sleep.DurationMore than 8 hours CGPA x More than 8 hours -0.011
## CGPA:Sleep.DurationOthers CGPA x Others 0.403
## Odds_Ratio Std_Error p_value
## CGPA 1.041 0.018 0.0224
## Sleep.Duration7-8 hours 1.184 0.187 0.3660
## Sleep.DurationLess than 5 hours 1.631 0.184 0.0080
## Sleep.DurationMore than 8 hours 0.856 0.193 0.4201
## Sleep.DurationOthers 0.036 2.603 0.2001
## CGPA:Sleep.Duration7-8 hours 0.992 0.024 0.7398
## CGPA:Sleep.DurationLess than 5 hours 0.978 0.024 0.3561
## CGPA:Sleep.DurationMore than 8 hours 0.989 0.025 0.6662
## CGPA:Sleep.DurationOthers 1.497 0.337 0.2310DT::datatable(summary_table, caption = "Odds Ratios and p-values for Sleep Duration and CGPA")barplot(summary_table$Odds_Ratio,
names.arg = summary_table$Term,
col = "skyblue",
main = "Odds Ratios by Sleep Duration and CGPA",
ylab = "Odds Ratio",
las = 2)
abline(h = 1, col = "red", lty = 2)
Results Interpretation
I conducted a logistic regression that examined the effects of academic performance (CGPA), sleep duration, and their interaction on depression status. I hypothesized that lower GPA would be associated with higher odds of depression, and that this relationship would be stronger with students with short sleep duration. However, contrary to my hypothesis, CGPA was significantly positively associated with depression, and higher GPA was linked with a slight increase in odds of depression. With sleep duration, students sleeping less than 5 hours had significantly higher odds of depression compared to the reference group of 5-6 hours. No other sleep duration categories differed significantly from the reference. The interaction terms between CGPA and sleep duration were not statistically significant, indicating the relationship between academic performance and depression did not vary significantly by sleep duration. This means my hypothesis that negative association between CGPA and depression would be stronger with students with shorter sleep was not supported.
Discussion
My findings that students who sleep less than 5 hours are at a significantly higher risk of depression aligns with prior research showing the impact of insufficient sleep on mental health (Berger et al., 2019). Similarly, Verboom et al. (2015) found that academic stress and lower academic achievement contribute to depression, which supports my findings of a complex relationship between CGPA and depression. However, the results also revealed that higher GPAs had slightly increased odds of experiencing depression. I theorize that this is due to heightened stress expirienced by the higher achieving students that is also consistent with prior research. The study’s cross-sectional design limits casual interpretations, which is also noted by Berger et al. (2019). Additionally, self-reported sleep and depression has a possibility of being biased. Verboom et al. (2015) also emphasized the need to consider factors of stress and social support in future research. In any future research, these factors should be included. The post-hoc power analysis showed that the study had good statistical power (1.0) to detect the observed effect size (0.235) given the large sample size. The high power suggests that the likelihood of a Type II error is extremely low, meaning, the significant association found between short sleep duration and depression is unlikely due to the insufficient sample size. In summary, my hypothesis that lower academic performance is associated with higher likelihood of depression, especially among students with shorter sleep, is only partially supported. Short sleep was significantly related to depression, however there was a positive association with higher CGPA and depression.
Sources
Berger, R. H., Diaz, A., Valiente, C., Eisenberg, N., Spinrad, T. L., Doane, L. D., Thompson, M. S., Hernández, M. M., Johns, S. K., & Southworth, J. (2019). The association between home chaos and academic achievement: The moderating role of sleep. Journal of Family Psychology, 33(8), 975–981. https://doi-org.dvc.idm.oclc.org/10.1037/fam0000535 Verboom, C. E., Sijtsema, J. J., Verhulst, F. C., Penninx, B. W. J. H., & Ormel, J. (2014). Longitudinal associations between depressive problems, academic performance, and social functioning in adolescent boys and girls. Developmental Psychology, 50(1), 247–257. https://doi-org.dvc.idm.oclc.org/10.1037/a0032547