Medical R: Why a Medical R Blog?

Peter Higgins

First and foremost, because a remarkable amount of (even published) medical research is not reproducible, meaning that, given the data and code, you should be able to reproduce all the published figures and tables.

Even better, we should have replicable research, meaning that someone else, following the same protocol, in a different setting, should be able to reproduce the results completely. This is a higher bar, and one we should strive for, but we are a long way from this.

As a first step toward reproducibility, we should collect data in a secure, ethical, and HIPAA-compliant way.

We should store data in standard (tidy) formats, and validate their quality during (not after) data entry.

We should have an audit trail of every change to the data, and a documented reason, date, and identity of the person making the data edit.

We should keep data out of spreadsheets, which do not promote reproducible data analysis, and frequently get in the way.

We should preserve (and comment) our data analysis in code, so that it can be re-run and reproduced in the future.

We should document our raw data, our code (all steps), and our computing environment, including all computer systems and software and packages used, so that others can reproduce our work.

Ideally, we should preserve a snapshot of our computing environment, with tools like Docker and binderhub to allow easy reproduction of our data analysis even years in the future.

If our data, and each step in the analysis, can not be reproduced, it should not be trusted.

Citation

For attribution, please cite this work as

Higgins (2020, Oct. 14). Medical R: Why a Medical R Blog?. Retrieved from https://higgi13425.github.io/medical_r/posts/2020-10-14-why-a-medical-r-blog/

BibTeX citation

@misc{higgins2020why,
  author = {Higgins, Peter},
  title = {Medical R: Why a Medical R Blog?},
  url = {https://higgi13425.github.io/medical_r/posts/2020-10-14-why-a-medical-r-blog/},
  year = {2020}
}