Skip to content

README.md

Latest commit

 

History

History
51 lines (36 loc) · 2.31 KB

README.md

File metadata and controls

51 lines (36 loc) · 2.31 KB

Gaming Addiction Study

Context

A large gaming-addiction dataset (~1% of the general population) collected by a YouTuber and former psychologist. The data is messy — many free-text responses were never formalized, and the core questionnaire was designed intuitively without scientific validation.

Main hypotheses

  1. Adolescents (13–17) and young adults (18–22) score higher on the questionnaire.
  2. Players who play more games per week score higher.
  3. Players with a high current ranking score higher.
  4. In the 18+ age groups, the relationship between games per week and engagement is stronger.

Secondary hypotheses

  1. Players can be clustered based on their attitudes toward the game.
  2. These clusters differ by age, engagement, hours played, and ranking.

Additional goals

  1. Compare frequentist and Bayesian approaches on the same dataset.
  2. Automate results formatting to APA standards.

Stack

R

  • tidyverse — data wrangling and pipelines
  • pwr — power analysis
  • emmeans — pairwise comparisons
  • rempsyc, broom, flextable, apaTables — APA-formatted output
  • poLCA — latent class analysis
  • ez — frequentist ANOVA
  • BayesFactor — Bayesian ANOVA

Workflow

  1. Data cleaning: missing values, outliers, impossible values, format normalization (Game_addiction_prep)
  2. Exploration: variable distributions, exploratory factor analysis, visualization (Game_addiction_main)
  3. Hypothesis testing: assumption checks, linear models, pairwise comparisons, visualization (Game_addiction_main)
  4. Post-hoc power analysis (Game_addiction_main)
  5. APA-formatted reporting (Game_addiction_main)
  6. Latent Class Analysis: converted items to 3-point scales, ran LCA to classify gaming attitudes (Game_addiction_main)
  7. Bayesian vs. frequentist comparison on secondary hypotheses (Game_addiction_main)

Results

  1. Hypotheses 1 & 2 confirmed: higher scores in adolescents/young adults and in heavy players.
  2. Hypotheses 3 & 4 not confirmed: ranking and age-moderated engagement effects were not significant.
  3. LCA produced interpretable clusters — the "aggressive teenager" stereotype has empirical backing.
  4. Bayesian and frequentist approaches converged on this large dataset, though Bayesian interpretation was more intuitive.