rtemis/prostate-cancer

Prostate cancer

Serum prostate-specific antigen (PSA) and eight clinical measures in 97 men with prostate cancer who were about to receive a radical prostatectomy (Stamey et al., 1989). The outcome is the log of PSA. The dataset is the prostate data of The Elements of Statistical Learning (ESL; Hastie, Tibshirani, and Friedman, 2009), with the book's train/test indicator, in a single Parquet file with one row per patient.

Structure

data/prostate-cancer.parquet: 97 rows, 11 columns. There are no missing values.

Column	Type	Description
id	string	Subject number, 1 to 97, in the order of the ESL file.
esl_train	logical	TRUE for the 67 training and FALSE for the 30 test cases of ESL Section 3.2.1.
lcavol	double	Natural log of cancer volume.
lweight	double	Natural log of prostate weight, in grams.
age	integer	Age, in years.
lbph	double	Natural log of the amount of benign prostatic hyperplasia.
svi	integer	Seminal vesicle invasion: 1 (present) or 0 (absent).
lcp	double	Natural log of capsular penetration.
gleason	integer	Gleason score.
pgg45	integer	Percentage of Gleason scores 4 or 5.
lpsa	double	Natural log of serum PSA.

lpsa is the last column. id and esl_train identify cases and are not predictors: drop them before training, or use esl_train to reproduce the ESL split. The dataset has one train split that holds all 97 patients.

Load

R - using rtemis (1.4.3 or later):

dat <- read("hf://datasets/rtemis/prostate-cancer")

R - using croissant:

library(croissant)
dat <- read_record_set(
  "hf://datasets/rtemis/prostate-cancer/croissant.json",
  "patients"
)

R - using nanoparquet:

url <- "https://huggingface.co/datasets/rtemis/prostate-cancer/resolve/main/data/prostate-cancer.parquet"
path <- file.path(tempdir(), "prostate-cancer.parquet")
download.file(url, path, mode = "wb")
dat <- nanoparquet::read_parquet(path)

Python - using polars:

import polars as pl
dat = pl.read_parquet("hf://datasets/rtemis/prostate-cancer/data/prostate-cancer.parquet")

Notes

• lbph and lcp take the value log(0.25) = -1.386 where the measure is 0: 43 values of lbph and 45 of lcp.
• gleason takes the values 6 (35 patients), 7 (56), 8 (1), and 9 (5).
• svi is 1 in 21 patients.
• Subject 32 carries the corrected prostate weight of 44.9 g, as in the ESL second edition.
• ESL standardizes the features to mean 0 and variance 1 before the analyses of Section 3.2.1 and beyond.

Source and processing

The file prostate.data of ESL was read without changes to the values. The unnamed row number column was named id. The train column (T or F) was stored as the logical column esl_train, placed after id.

Metadata

croissant.json is the dataset's Croissant 1.1 metadata, written with the croissant R package. It describes the file with its SHA-256 checksum, each field, and the Responsible AI properties of the data.

License

The source file and the publication state no license. The data are distributed by the ESL authors for use with the book, and are reproduced here for teaching.

Citation

Cite the study that collected the data and the book that distributes them:

@article{stamey1989prostate,
  author = {Stamey, Thomas A. and Kabalin, John N. and McNeal, John E. and Johnstone, Iain M. and Freiha, Fuad and Redwine, Elise A. and Yang, Norman},
  title = {Prostate specific antigen in the diagnosis and treatment of adenocarcinoma of the prostate. {II}. Radical prostatectomy treated patients},
  journal = {The Journal of Urology},
  year = {1989},
  volume = {141},
  number = {5},
  pages = {1076--1083},
  doi = {10.1016/s0022-5347(17)41175-x}
}

@book{hastie2009elements,
  author = {Hastie, Trevor and Tibshirani, Robert and Friedman, Jerome},
  title = {The Elements of Statistical Learning: Data Mining, Inference, and Prediction},
  edition = {2},
  publisher = {Springer},
  address = {New York},
  year = {2009},
  doi = {10.1007/978-0-387-84858-7}
}