Datasets from Computer Age Statistical Inference.

CASIdata

CASIdata provides the datasets from Efron & Hastie (2016, ISBN: 9781108107952), Computer Age Statistical Inference: Algorithms, Evidence, and Data Science in an accessible R format for those who want to use them for teaching, study or to try to reproduce or extend analyses from the book. They were downloaded from Trevor Hastie’s web site, https://hastie.su.domains/CASI_files/DATA/, but quite a few files were messy and required some processing to make into R datasets.

Even so, some of the datasets may require data cleaning, renaming of variables, re-shaping or other tidying steps to be useful for analysis. But that’s part of learning.

Installation

This package is not yet on CRAN. You can install it from this GitHub repo or from R-universe

remotes::install.github("friendly/CASIdata")
install.packages('CASIdata', repos = c('https://friendly.r-universe.dev'))

Datasets included here

Loading package: CASIdata

Missing Datasets

The following dataset appears in data-raw/CASI-save.R but is not (yet) included in the package:

See data-raw/missing-datasets.md for details on resolving this.

External Datasets (Not Included)

These large datasets are referenced in the book but not included in the package due to size constraints. They can be downloaded directly from the sources listed below.

CASI datasets (too large for CRAN)

• protein_kernel: 1708 x 1708 inner-product (kernel) matrix for human proteins (Section 19.6). Computed using a string kernel on bag-of-4-grams amino acid representations.
  • Source: https://hastie.su.domains/CASI_files/DATA/protein_kernel.txt
  • Load in R: protein_kernel <- matrix(scan("https://hastie.su.domains/CASI_files/DATA/protein_kernel.txt", what=0), 1708, 1708)
• protein_label: Response labels (-1/+1) for the 1708 proteins (45 positives, 1663 negatives).
  • Source: https://hastie.su.domains/CASI_files/DATA/protein_label.txt
  • Load in R: protein_label <- scan("https://hastie.su.domains/CASI_files/DATA/protein_label.txt", what=0)
• prostmat: 6033 x 102 gene expression matrix comparing 50 controls vs 52 prostate cancer patients (Section 3.3).
  • Source: https://hastie.su.domains/CASI_files/DATA/prostmat.csv
  • Load in R: prostmat <- read.csv("https://hastie.su.domains/CASI_files/DATA/prostmat.csv")
  • Note: Column names need cleanup (see data-raw/missing-datasets.md for renaming code)
• leukemia_big: 7128 x 72 gene expression matrix (10MB). A larger version of leukemia_small.
  • Source: https://hastie.su.domains/CASI_files/DATA/leukemia_big.csv
  • Load in R: leukemia_big <- read.csv("https://hastie.su.domains/CASI_files/DATA/leukemia_big.csv")

Image datasets (hosted externally)

• CIFAR-100: 100 image classes, 600 images each (32x32x3 color). Used in Chapter 18.
  • Source: https://www.cs.toronto.edu/~kriz/cifar.html
• MNIST: Handwritten digit database, 60K training + 10K test images (28x28 grayscale). Used in Chapter 18.
  • Source: http://yann.lecun.com/exdb/mnist/

Variable Renaming

Some datasets had variables renamed for clarity:

Example

No examples yet.

library(CASIdata)
## basic example code