wmda_mv4

WMDA matching validation code

Requirements

This system works from a config file. The default location is conf/config.json

The data sources are not included in this repo.

MV4 dataset

The MV4 dataset needs to be downloaded. A good location is data/MV4. The files donors.csv and patients.csv should be in that directory.

results datasets

The results datasets should be downloaded.

A good location is data/results/REGISTRY/DATE/results.csv.

For the comparison, a single letter code is used to identify this registry in the config file and will refer to it in the output.

parameters

keep_zero_prob

If this is set to 0 then any cases where P(8/10) + P(9/10) + P(10/10) <=0 will be excluded from analysis. Note the <= is needed because at least one participant saw fit to use negative probability values.

Procedure

1. download MV4 dataset and place it in data/MV4/. The expected files are patients.csv and donors.csv.
2. set the mv4_datadir variable in config.json approrpriately
3. download results and place them in data/results/REGISTRY/DATE/
4. set update config.json results list with an apporpriate letter and filepath
5. repeat the previous two steps for all participating groups
6. run scripts/joinall.pl
7. output will be in output/all.txt
8. run scripts/stats.pl
9. output will be in `output/stats.txt'

categorize

1. preformat scripts/preformat_d.pl >data/donors.f.csv

2. preformat scripts/preformat_p.pl >data/patients.f.csv

3. pyard-reduce-csv -c conf/donor.reduce_conf.json -i 3.52.0

4. pyard-reduce-csv -c conf/patient.reduce_conf.json -i 3.52.0

5. scripts/mkfilter.pl >data/patients.filter.csv