Sample Size for Validation of Risk Models with Binary Outcomes

Estimation of the required sample size to validate a risk model for binary outcomes, based on the sample size equations proposed by Pavlou et al. (2021) . For precision-based sample size calculations, the user is required to enter the anticipated values of the C-statistic and outcome prevalence, which can be obtained from a previous study. The user also needs to specify the required precision (standard error) for the C-statistic, the calibration slope and the calibration in the large. The calculations are valid under the assumption of marginal normality for the distribution of the linear predictor.

output: html_document: df_print: paged word_document: default pdf_document: default

sampsizeval

The purpose of sampsizeval is to perform sample size calculations for the validation of risk models for binary outcomes.

Installation

The development version can be installed from GitHub with:

# install.packages("devtools")
devtools::install_github("mpavlou/sampsizeval")

Example

This is an example of a sample size calculation to validate a risk model for a binary outcome. The anticipated values of the outcome prevalence and the C-statistic are p=0.1 and C=0.75, respectively.

library(sampsizeval)

The target is to calculate the size of the validation data so as to estimate the C-statistic, the Calibration Slope and the Calibration in the Large with sufficient precision. In this example, the required precision is reflected by a SE of the estimated C-statistic of at most 0.025, and SE of the estimated Calibration Slope and Calibration in the Large of at mos 0.1.

sampsizeval(p=0.1, c=0.75, se_c=0.025, se_cs =0.1, se_cl = 0.1)

The recommended sample size is 1536 observations.