DATA1001 | University of Sydney | Semester 2, 2026
International students spend notably more time on social media daily (median 5.0 hrs) than domestic students (median 3.5 hrs). However, social media use is a very weak predictor of intended study hours (R² = 0.010). Universities designing student support programs should look beyond screen time and consider confounding factors such as study habits and course load.
library(tidyverse)
# load data
data <- read.csv("DATA1001_Project_1.csv")
#clean data
clean <- data %>%
filter(consent == "I consent to take part in the study") %>%
filter(social_media_use >= 0 & social_media_use <= 24) %>%
filter(hours_studying > 0 & hours_studying <= 60)The data was sourced from a voluntary survey examining 29 variables, completed by 5,057 consenting DATA1X01 students across cohorts 2024S2 to 2026S2.
Variables focused on:
While the DATA1X01 dataset was collected from students enrolled in the unit, participation was voluntary, meaning the dataset represents a sample rather than the full population. The survey is self-reported, so students may inaccurately estimate their daily social media use and intended study hours. Social media use was measured via phone screen time only, potentially excluding desktop usage. As the data is cross-sectional, causation cannot be inferred from any observed relationships.
We assumed that students answered the survey honestly and that phone screen time is a reasonable proxy for social media use. Lecture and tutorial hours were assumed to be excluded from intended study hours. Values for social media use greater than 24 hours per day were treated as data entry errors and removed. Values of hours studying equal to 0 were excluded as non-meaningful, and values exceeding 60 hours per week were treated as unrealistic outliers and removed.
All 5,057 responses indicated consent, so no filtering was required on that basis. 43 responses with social media use greater than 24 hours per day were removed as implausible. Responses with hours studying equal to 0 or greater than 60 were excluded, resulting in a final cleaned dataset of 4,892 rows used for analysis
Do domestic and international students differ in their daily social media use?
library(tidyverse)
library(ggplot2)
data <- read.csv("DATA1001_Project_1.csv")
#filter for RQ1
rq1_data <- data %>%
filter(consent == "I consent to take part in the study") %>%
filter(social_media_use >= 0 & social_media_use <= 24) %>%
filter(student_type %in% c("Domestic", "International"))
#boxplot
ggplot(rq1_data, aes(x = student_type, y = social_media_use, fill = student_type)) +
geom_boxplot(outlier.colour = "grey50", outlier.size = 0.8, alpha = 0.8) +
scale_fill_manual(values = c("Domestic" = "#4E79A7", "International" = "#F28E2B")) +
labs(
title = "Daily Social Media Use by Student Type",
x = "Student Type",
y = "Social Media Use (hours/day)",
fill = "Student Type"
) +
theme_minimal() +
theme(legend.position = "none")
The boxplot shows that international students have a notably higher median daily social media use (5.0 hrs, IQR = 4.0) compared to domestic students (median 3.5 hrs, IQR = 3.1). Both distributions are right-skewed, with the international group showing a wider spread and more outliers, suggesting greater variability in that group. The greater variability among international students (SD = 3.03 vs 2.64) may reflect their reliance on social media to maintain connections with family and friends overseas (Schwartz & Omori, 2024). As this is an observational study, we cannot conclude that being an international student causes higher social media use — confounding variables such as cultural background, living situation, and time zone differences may also play a role.
Is there a linear relationship between daily social media use and intended study hours per week?
# scatter plot
ggplot(rq2_data, aes(x = social_media_use, y = hours_studying)) +
geom_jitter(alpha = 0.2, colour = "#4E79A7", size = 0.6, width = 0.2, height = 0) +
geom_smooth(method = "lm", se = TRUE, colour = "#E15759", formula = y ~ x) +
labs(
title = "Social Media Use vs Intended Weekly Study Hours",
x = "Social Media Use (hours/day)",
y = "Intended Study Hours per Week"
) +
theme_minimal()
library(plotly)
# residual plot
p <- ggplot(model, aes(x = .fitted, y = .resid)) +
geom_point(alpha = 0.3, colour = "#4E79A7", size = 0.6) +
geom_hline(yintercept = 0, linetype = "dashed", colour = "#E15759") +
labs(
title = "Residual Plot",
x = "Fitted Values",
y = "Residuals"
) +
theme_minimal()
ggplotly(p)The fitted linear model is: hours_studying = 4.48 + 0.18 × social_media_use (n = 4,892, R² = 0.010, r = 0.10).
The positive slope is actually counter-intuitive — we might expect higher social media use to reduce study hours, not increase them. However, with R² = 0.010 and r = 0.10, this relationship is extremely weak and should not be interpreted causally in either direction; the true correlation is close enough to zero that the direction of the slope may not be reliable or meaningful. The effect size is also trivial: even at 20 hours of daily social media use, the model only predicts 3.6 more study hours per week than a non-user, a negligible difference given study hours range from 0–60 in this dataset. The residual plot shows no strong curve, so linearity isn’t badly violated, but the wide, uneven spread of residuals (mild heteroscedasticity) confirms the model has little real predictive power. This suggests social media use alone is a poor predictor of study intentions, consistent with Amez & Baert (2020), who found the social media–academic effort relationship is largely mediated by individual self-regulation rather than screen time itself. A linear model using this single variable is not an adequate tool for explaining intended study hours, and the counter-intuitive direction of the slope further reinforces that no meaningful causal relationship should be assumed from this analysis.
Schwartz & Omori (2024) found that international students use social media primarily to manage homesickness and maintain overseas connections, supporting our finding that international students report higher daily social media use than domestic students.
Amez & Baert (2020) found that the relationship between screen time and academic effort is largely mediated by self-regulation, challenging our weak positive slope and suggesting social media use alone cannot predict study intentions.
References
Schwartz, M., & Omori, K. (2024). Communities of Practice and Acculturation: How International Students in American Colleges Use Social Media to Manage Homesickness. Journal of International and Comparative Education, 13(1), 57–72.
Amez, S., & Baert, S. (2020). Smartphone use and academic performance: A literature review.
Group Meetings:
12/08/2026
All member attended
09/09/2026
All member attended
11/09/2026
All member attended
13/09/2026
All member attended
16/09/2026
All member attended
Contributions:
Leo
Clean the dataset, identifying and removing 43 social media outliers and excluding unrealistic study hour response
Conducted RQ1 analysis
Built the RQ2 linear model and interpreted the slope, R² and residual plot
Created all visualisations including an interactive residual plot using plotly in the HD examplar
Implemented tabbed panel layout for RQ2 plots
Formatted the final report using Quarto with flatly theme, table of contents and code folding
Debugged rendering errors including working directory issues, YAML formatting errors and chunk syntax errors (eg. include=FALSE, message=FALSE, setwd() )
Wrote all written analysis including EDA,RQ1 and RQ2 interpretations,limitations, assumptions and data cleaning sections
Give crafting presentation slide instruction
Tanay
reviewed and proofread the written analysis for clarity and accuracy
contributed to selecting the research questions
Reviewed data cleaning decisions and provided feedback on outlier removal
Crafting Presentation
Manthan
Assisted in interpreting the RQ1 boxplot findings
Reviewed the executive summary and suggested edits
Contributed to the limitations and assumptions discussion
Crafting Presentation
Haoxuan
Reviewed the RQ2 linear model interpretation
Assisted in selecting appropriate variables for analysis
Proofread the final before submission
Give report analysis
Crafting presentation
Anoushka
Assisted in interpreting the RQ1 boxplot findings
Reviewed the executive summary and suggested edits
Contributed to the limitations and assumptions discussion
Crafting presentation
Dhairya
Reviewed the RQ2 linear model interpretation
Assisted in selecting appropriate variables for analysis
Proofread the final before submission
Crafting presentation
Resources Used:
Data1001 project 1 Rubric
Data 1001 R Template
Data1001 project 1 HD examplar
Tutor Guidance (Kevin, Anina)
R documentation, ggplot2,tidyverse,plotly package documentation
Quarto documentation, for tabset panel layout and YAML formatting
Data 1001 lecture notes for statistical terminology (observational study, confounding variables, linearity assumption and heteroscedasticity)
AI Usage Statement:
Claude, Anthropic, https://claude.ai was used to assist in planning the structure of this report, selecting appropriate research questions from the dataset, computing preliminary statistics, and drafting the written analysis. All R code was written and verified by the students. All statistical interpretation and conclusions are the group’s own.
Shared Values and Ethical Principles:
Shared Value — Integrity:
We ensured that our analysis was directed by the data rather than desired outcomes. Outliers were removed transparently with documented reasoning, and we reported findings including the weak R² without exaggerating the strength of the relationship.
Ethical Principle — Respect:
We recognised that the survey data involves real students’ personal information. We did not attempt to identify individuals and handled sensitive variables (e.g., drug_use) with care, choosing not to include them in our analysis.