Assessing the performance of prediction models with binary outcomes: a framework for some traditional and novel measures
R Code repository for the manuscript 'Assessing the performance of prediction models: a framework for some traditional and novel measures' by Steyerberg et al. (2010). We provided how to develop and validate a risk prediction model for binary outcomes using logistic regression including more recent updates about performance assessment (e.g. using calibration hierarchy definitions by Van Calster et al. (2016) here) .
The repository contains the following code:
-
Minimal and essential code to develop and validate a risk prediction model with binary outcomes when both development and validation data are available. People with basic or low statistical knowledge and basic R programming knowledge are encouraged to use these files. To reproduce the main results of the manuscript, this script is sufficient. A "quick and I hope not so dirty" Python code is here to reproduce the main results of the manuscript. I encourage users to install and use Python in RStudio as suggested, for example, here using PyCharm to install and update Python packages. In Rstudio is important to set up the Python interpreter through Rstudio global options.
-
Minimal and essential code to validate a risk prediction model in a external data when model equation of a developed risk prediction model is available. A more extensive and elaborated output is here and the corresponding .Rmd source code is here.
-
Extensive output and code to develop and validate a risk prediction model with a binary outcome. The .Rmd source code is here. People with advanced knowledge in statistics are encouraged to use these files. The corresponding "experimental" extensive and more elaborated Python code developed in RMarkdown is here (still work in progress). The .Rmd source code mostly based on Python code is here.
External functions and figures are available in the corresponding subfolders.
You can either download a zip file containing the directory, or you can clone it by using
git clone https://github.com/danielegiardiello/ValRegMod.git
In either case, you can then use the ValRegMod.Rproj
file to open
and Rstudio session in the directory you have just downloaded. You may then knit
both rmarkdown files, or run them line-by-line.
Name | Affiliation | Role |
---|---|---|
Daniele Giardiello | University of Milan-Bicocca (IT) | Author/maintainer |