SM-2 vs. Cognitive-Load Adaptive Spaced-Repetition

R simulations & analysis comparing the classic SM-2 scheduler with a cognitive-load–adaptive variant.

---

📋 Overview

This project implements two spaced-repetition schedulers in R:

1. SM-2 (SuperMemo-2): the standard ease-factor algorithm
2. CLoad: an adaptive variant that adjusts intervals based on simulated “cognitive load”

We run synthetic cohort simulations (N users × M cards) and analyze:

• Day-8 recall accuracy (t-test & Cohen’s d)
• Learning curves over days (mixed-ANOVA)
• Self-efficacy pre/post comparisons
• Retention curves (ggplot2)

---

📂 Folder Structure

├── README.md
├── remindXR-study.Rproj
├── data/
│   ├── cards.csv
│   ├── day8.csv
│   ├── pt.csv
│   ├── logs.csv
│   ├── selfeff.csv
│   ├── def_pt.rds
│   └── all_analysis_data.RData
├── R/
│   ├── schedulers.R          # SM-2 & CLoad functions
│   └── simulate_cohort.R     # simulate_study() wrapper
├── analysis/
│   ├── analysis.Rmd          # statistical tests & plots
│   └── analysis.html         # knitted output

---

🚀 Getting Started

Prerequisites

• R (≥ 4.0)
• RStudio (recommended)
• Required packages:

  install.packages(c(
    "dplyr", "magrittr", "simstudy",
    "knitr", "rmarkdown", "ggplot2", "lme4", "effsize"
  ))
  

---

Simulation

1. Open the project file remindXR-study.Rproj in RStudio.

2. Source the scheduler functions and simulate:

   source("R/schedulers.R")
   source("R/simulate_cohort.R")
   
   cards_df <- read.csv("data/cards.csv")
   logs <- simulate_study(cards_df, days = c(0,1,3,5,7,8))
   head(logs)
   

Analysis

1. In the analysis/ directory, open analysis.Rmd.
2. Knit to HTML or PDF to reproduce all:
   • Day-8 t-test & Cohen's d
   • Mixed-ANOVA on accuracy trends
   • Self-efficacy pre/post Wilcoxon test
   • Retention curves (ggplot2)
3. View the finished report at analysis/analysis.html

---

📊 Data

• cards.csv: initial deck definitions
• day8.csv: summary recall percentages
• logs.csv: long-form review logs (all days)
• selfeff.csv: self-efficacy pre/post data
• pt.csv: participant metadata
• def_pt.rds, all_analysis_data.RData: serialized R objects for quick load

---

🛠️ Structure & Extensibility

• R/schedulers.R:
  • sm2(ease, q, interval_prev)
  • cload(ease, q, load, interval_prev)
• R/simulate_cohort.R:
  • simulate_study(cards_df, days) drives the day-by-day review loop
• analysis/analysis.Rmd:
  • Organize code chunks for reproducible figures and tables

---

📄 License

This work is licensed under the MIT License. For details, see the LICENSE.