Missingness quantification

This repository performs missing data analysis on data at different sites and reports the results in the results/ folder.

Please run quantify-missingness.Rmd, predict-TE.Rmd, and table-one.Rmd and send us your results!

Repository structure

Rmd files

• quantify-missingness.Rmd uses the {naniar} package to quickly generate figures of missing data information (avg proportion missing, demographic stratification, temporal analyses).
• predict-te.Rmd performs LDA on matrix of number of valid lab values for each patient for the first 10 days of hospitalization and identifies correlation between each topic's value with the outcome (TE, AKI, Severity, Neuro).
• table-one.Rmd will likely be used for descriptive papers for generating Table 1 of demographic statistics of patients with and without thrombotic events.

Other files

• htmls/ contains rendered html reports.
• old/ contains old exploratory scripts.
• R/ contains utility scripts such as for processing, mapping ICD codes to comorbidity, summary statatistics and other utility functions.

Which script should I run?

The best way to run this analysis is to clone this repository on your local machine (please ensure you're in a directory where you want the repository to be downloaded):

git clone https://github.com/ameliatlm/missing-data-4ce.git

Then, go inside the repository:

cd missing-data-4ce

and make a copy of quantify-missingness.Rmd, name it with your site name, for example:

cp quantify-missingness.Rmd quantify-missingness-penn.Rmd

and open the R project

open missing-data-4ce.Rproj

and navigate to the newly created file (e.g. quantify-missingness-penn.Rmd) to modify the code to run on the data at your specific site.

All you must do is change the params at the beginning of the .Rmd file. data_dir refers to the directory where your site's data is located, package_dir refers to the directory where the missing-data-4ce folder is located from cloning the package, dateFormat refers to the date format that your site uses, results_file should contain the name of your site instead of "penn", and site should be changed to the name of your site instead of "penn".

Once everything runs, please hit the "Knit" button on top of the .Rmd file to create an .html file that will automatically be put into htmls/.

Finally, please upload your results (in results/ and htmls/) via a pull request or request @ameliatlm to add you as a contributor.

Please also repeat these instructions for predict-TE.Rmd and table-one.Rmd! Starting with making a copy of the .Rmd file.

If you run into any problem adapting this code to your data, let us @ameliatlm know via Slack or submit an issue.