Prelaunch preview. Explore AROS products and skills. Checkout is in sandbox; no real payment.
← CohortDAG documentation

Reproduce a published study end to end: Russell et al. 2020

Tutorial · CohortDAG 0.2.9

Find the study and get its data

Skip the first-run tour

When CohortDAG opens for the first time, a short tour introduces the window. Click Skip to close it for now. You can reopen it at any time from the Help button.

Skip the first-run tour

Find the study in Pipelines

Click Pipelines. Under Studies, CohortDAG lists published papers that it reproduces. Find the 2020 Diamond Princess study by Russell and colleagues. It estimates COVID-19 fatality on board the cruise ship.

Find the study in Pipelines

Get the data

The study needs three public datasets that are not on this computer yet. Click Get data for each one. CohortDAG downloads every file and checks it against its recorded checksum before marking the data ready.

Get the data

Load the pipeline and arrange the graph

Load and arrange the pipeline

Click Load. The pipeline opens on the canvas, with the results of an earlier run already restored. Close the results panel to give the graph room, then click Arrange Graph. Each data source moves to the top, and its steps flow down to the estimate they feed.

Load and arrange the pipeline

Read the method in the node guide

The estimate comes from a Delay-Adjusted CFR node. Open the Nodes tab and click the info button next to Delay-Adjusted CFR. Its guide explains the delay-adjusted method and how to read the numbers.

Read the method in the node guide

Run it and read the result

Run the pipeline

Press Escape to close the guide, then click Run All Outputs. CohortDAG compiles the pipeline and runs it on Apache Spark, reusing results already computed from the same data. When the run finishes, the results panel opens. Close it to keep the whole graph in view. The log on the right records each step.

Run the pipeline

Read the result

Open the Workspace tab to read the selected result as a table. In the adjusted row, the delay-adjusted case fatality ratio is 0.026, or 2.6%. In the ifr row, the infection fatality ratio is 0.013, or 1.3%. These are the values the paper reports.

Read the result