Skip to contents

Example Data for Conditional Logistic Regression

ExampleData_cc.Rd

A simulated dataset generated for 1:M matched case-control studies (Conditional Logistic Regression, CLR). The data is organized into matched sets (strata), with exactly one case (y=1) and \(m=4\) controls (y=0) per set.

Usage

data(ExampleData_cc)

Format

A list containing the following elements:

train

A list with components for training the CLR model:

z

Numeric matrix of covariates (dimension \(n_{\mathrm{train}}\times 6\)) with columns named Z1–Z6.

y

Binary outcome vector (1=case, 0=control).

stratum

Integer vector identifying the matched set for each observation (200 unique strata in train).

test

A list with the same structure as train, used for external evaluation (500 unique strata in test).

beta_external

Numeric vector (length 6) of CLR coefficients estimated on a separate external dataset using all Z1–Z6.

Details

Data-generating mechanism:

  • Study design: 1:4 matched case-control study (\(m=4\) controls per case).

  • Covariates: 6 variables (Z1–Z6) drawn from a correlated multivariate normal distribution.

  • True coefficients: \(\beta = (1, -1, 1, -1, 1, -1)\).

  • Set-specific effect: A random stratum-specific intercept \(\theta_i \sim N(0, 0.5^2)\) is added to the linear predictor; it is eliminated by CLR conditioning.

  • Outcome generation: Within each stratum \(i\), the single case (y=1) is selected with probability proportional to \(\exp(\theta_i + Z^\top \beta)\).

  • External beta estimation: beta_external is obtained by fitting ncc on a separate simulated dataset with a slightly different true coefficient vector \(\beta_{\mathrm{ext}} = (0.8, -0.8, \dots)\) and correlation \(\rho = 0.3\), using the "breslow" tie approximation.

Examples

data(ExampleData_cc)