2026 ASA South Florida Student Data Challenge

Dataset Description

The dataset includes 1,200 individuals and 97 variables sampled from the 2024 National Health and Nutrition Examination Survey (NHANES).

The outcome variable is:

• LBDHDD_outcome – a noise-perturbed version of Direct HDL-Cholesterol (mg/dL). This prevents reconstruction of the original NHANES values while preserving realistic structure.

Predictors include:

• Dietary intake from the 24-hour dietary recall interview
• Demographics (age, gender, race/ethnicity, income-to-poverty ratio, marital status)
• Anthropometrics (BMI, waist circumference)
• Alcohol indicators and diet-behavior variables

Training: 1,000 observations
Test: 200 observations

All variables retain NHANES labels and can be inspected using attr(x, "label").

Technical resources: Data files and official competition documents are hosted in the GitHub repository .

Downloading the Data in R

### Download Training Data

tmp <- tempfile()
download.file("https://luminwin.github.io/ASASF/train.rds", tmp, mode = "wb")
train <- readRDS(tmp)

### View Variable Labels

lapply(train, attr, "label")

### Download Test Data

download.file("https://luminwin.github.io/ASASF/test.rds", tmp, mode = "wb")
test <- readRDS(tmp)

You can also download the CSV files (train.csv, test.csv, variable_labels.csv) directly from the GitHub repository https://github.com/luminwin/ASASF by clicking Code → Download ZIP.

Eligibility

• Any student is eligible to participate, including high school, undergraduate, and graduate students, with no geographic restrictions.
• Submissions may be made by an individual or a team of up to 4 members.
• Each participant or team is allowed one submission.
• For team submissions, the competition level is determined by the highest academic level among all team members. Mixed-level teams will be evaluated in the category corresponding to that highest level.

Competition Tasks

1. Prediction Track (High School, Undergraduate, and Graduate)

Predict LBDHDD_outcome for the test dataset.

Submission Requirements

• A CSV file named pred.csv
  • Exactly one column, named pred
  • The number of rows must match the test dataset
  • Predictions must be in the same order as the test data
• A report (maximum 4 pages) named <participant_or_team_name>.pdf, summarizing:
  • Model and preprocessing choices
  • Validation and/or tuning strategy
  • The final model used to generate predictions, including code snippets or a link to the full code repository

Allowed Software: Participants may use any software, programming language, or computational tools to generate their submissions.

Evaluation Process

• Submissions will first be ranked based on Root Mean Squared Error (RMSE) evaluated on the test dataset.
• The top 30% of participants within each competition level (high school, undergraduate, graduate) will advance to the final evaluation stage.
• In the final stage, submissions will be reviewed by judges and ranked based on the quality, clarity, and rigor of the submitted reports.

2. Visualization Track (High School)

Submit a 1–4 page PDF containing:

• At least two clear visualizations based on variables in the training dataset
• Short explanations describing patterns, trends, or insights

No modeling is required for this track.

Submission

Submit the following files:

• pred.csv (prediction track only)
• participant_or_team_name.pdf
• Optional: source code or notebook files

Submissions should be uploaded through the designated form. (File uploads require Google login; this is a Google Forms requirement.)

Award

• $150 per winner, with four winners total:
  – High School (Prediction Track)
  – High School (Visualization Track)
  – Undergraduate (Prediction Track)
  – Graduate (Prediction Track)
• Winners will be recognized and invited to present posters at the 2026 Annual ASA Florida Chapter Meeting.
• The top 20% of participants will receive a certificate of recognition.
• The top 10% of participants will receive a waived registration fee for the 2026 Annual ASA Florida Chapter Meeting.

Awards will be issued as cash prizes to U.S. participants. For international participants, awards will be provided as reimbursement for eligible expenses (e.g., registration or travel), in accordance with ASA administrative and tax guidelines.

Result Announcements

• 02/23/2026 — RMSE results and overall quantile rankings were sent to all prediction track teams. As reflected in the attached filename (e.g., …RMSE_####-XXth-to-XXth.csv), each team received an individual RMSE score and percentile ranking among all participants. The attached CSV file contains two columns: the team's submitted predicted values and the observed outcomes from the test dataset.
• 02/25/2026 — All award-winning teams were notified and received a digital award certificate. Teams that did not receive an award were also informed of their results. Teams whose submitted reports received written feedback were copied on that feedback. Award levels were primarily determined by RMSE ranking, unless a team's reviewer score on the submitted report was exceptionally low. Four teams were invited to present at the 2026 Annual ASA Florida Chapter Meeting. Cash award winners were given first priority; if a winning team was unable to attend in person, the next available team with the lowest RMSE was invited in their place.

Cash Prize Winners

• Graduate: Jade Muyambo — University of Miami (Mentor: Dr. Liang Liang) RMSE: 4.4336, Rank 1st Overall (Presented on site)
• Undergraduate: Hayden Jose — University of Central Florida (Mentor: Dr. Maria Capursi) RMSE: 4.4611, Rank 3rd Overall
• High School (Prediction): Mandy Wu, Kyra Movva, Daniel Cai — American Heritage Broward RMSE: 4.6331
• High School (Visualization): Jacci Zhang (Ransom Everglades Upper School) & Arya Salem (Seven Lakes High School, Texas) (Mentor: Klara Hankollari)