Skip to content

Data Card: Palmer Penguins

This Data Card documents the dataset used by the penguins-body-mass experiment.

It follows the general transparency goals of Google's Data Cards Playbook: describe dataset provenance, composition, intended use, limitations, and considerations not apparent from the data.

Dataset Summary

Item Description
Dataset Palmer Penguins
Curated dataset penguins
Observations 344 penguins
Species Adelie, Chinstrap, Gentoo
Location Palmer Archipelago, Antarctica
Islands Biscoe, Dream, Torgersen
Study period 2007-2009
Grain one penguin
Primary use here supervised regression
Target in this experiment body_mass_g
Selected feature in this experiment flipper_length_mm

Purpose

The Palmer Penguins dataset provides measurements and descriptive attributes for penguins observed in the Palmer Archipelago.

The curated dataset was designed as an accessible dataset for data exploration and visualization and is commonly used as an alternative to the Iris dataset.

Provenance

The underlying observations were collected by Dr. Kristen Gorman and the Palmer Station Long Term Ecological Research program.

The palmerpenguins project made curated versions of the data readily available for teaching, exploration, and analysis.

This project obtains the dataset through Seaborn's penguins dataset interface.

Dataset Composition

The dataset contains 344 observations representing individual penguins.

The variables available through the Seaborn version used in this project are:

  • species
  • island
  • bill_length_mm
  • bill_depth_mm
  • flipper_length_mm
  • body_mass_g
  • sex

The dataset includes three penguin species:

  • Adelie
  • Chinstrap
  • Gentoo

Missing Data

Some observations contain missing values.

For this experiment, only two columns are required:

  • flipper_length_mm
  • body_mass_g

Two of the 344 observations are missing one of these required values.

The declared experiment policy drops those observations, leaving:

342 modeling observations

No values are imputed.

Intended Use

The dataset is appropriate for:

  • education
  • exploratory data analysis
  • visualization
  • introductory statistical analysis
  • supervised machine-learning experiments
  • demonstrating reproducible analytical workflows

In this repository, the dataset is used to demonstrate a clear baseline-versus-candidate regression experiment.

Additional Exploration

Other reasonable analytical questions include:

  • predicting penguin species
  • predicting body mass from multiple morphological measurements
  • comparing measurements across species
  • examining differences among islands
  • studying relationships among bill dimensions, flipper length, and body mass

Those are separate analytical experiments and should have their own declared assumptions, selected features, evaluation methods, and conclusions.

Limitations

The dataset is small and represents penguins observed in a specific geographic region and study period.

Results should therefore not automatically be generalized to:

  • all penguin species
  • all geographic populations
  • different ecological conditions
  • future populations
  • other biological species

Measurements also contain missing values, and some variables may be associated with species, sex, island, or other biological structure.

A predictive relationship observed in this dataset should not be interpreted automatically as a causal relationship.

Representation Considerations

The dataset contains observations from three species and three islands, and those groups are not necessarily represented equally.

Model performance calculated across the complete held-out sample may therefore hide differences in performance across species, sex, or island.

A more advanced experiment could evaluate those groups separately.

Experiment-Specific Use

The feature choice is intentionally constrained. This repository uses only:

flipper_length_mm → body_mass_g

The purpose is to determine whether one interpretable morphological feature provides useful predictive information beyond a mean-value baseline.

Project Data Processing

The project:

  1. loads the Palmer Penguins dataset
  2. observes the available columns
  3. validates the selected feature and target
  4. selects flipper_length_mm and body_mass_g
  5. drops observations missing either required value
  6. performs the declared train/test experiment

References

◄ Back to Home