59 lines
3.5 KiB
Markdown
59 lines
3.5 KiB
Markdown
# shinyDAG
|
|
|
|
shinyDAG is a web application that uses R and LaTeX to create publication-quality images of directed acyclic graphs (DAGs). Additionally, the application leverages complementary R packages to evaluate correlational structures and identify appropriate adjustment sets for estimating causal effects<sup>1-4</sup>. The web-based application can be accessed at [https://apps.gerkelab.com/shinyDAG/](https://apps.gerkelab.com/shinyDAG/).
|
|
|
|
## Key operations
|
|
|
|
### Adding nodes and edges
|
|
|
|
|
|
|
|
### Editing DAG aesthetics
|
|
|
|
|
|
|
|
## Examplary usage
|
|
|
|
The following DAG was reproduced from "A structural approach to selection bias"<sup>5</sup> (Figure 6A) using the shinyDAG web app.
|
|
|
|
|
|
|
|
For comparison, the DAG from the original article is shown below.
|
|
|
|
|
|
|
|
The DAG represents a study on the effects of antiretroviral therapy (E) on AIDS risk (D), where immunosuppression (U) is unmeasured. L represents presence of symptoms (such as fever, weight loss, and diarrhea) and C represents censoring. A spurious path exists between E and D due to selection bias. We can see this in shinyDAG by ensuring that we've selected E as the exposure, D as the outcome, adjusted for C, and then toggling the "Examine DAG elements" button in the bottom left corner. The spurious open path is displayed as D <- U -> L -> C <- E.
|
|
|
|
|
|
|
|
One possible resolution for this bias is to adjust for L. After toggling L in the "Select nodes to adjust" section, we see that all spurious E to D paths are now closed.
|
|
|
|
|
|
|
|
## Other features
|
|
|
|
In addition PDF and PNG exports, users can download R objects in `ggdag` or `daggity` formats, as well as the source LaTeX code. The "Edit LaTeX" pane permits in-app modification of the LaTeX code with a preview window; however, users should be aware that the information in "Examine DAG elements" is not responsive to changes in the Edit LaTeX pane.
|
|
|
|
shinyDAG should work in most modern web browsers, however, we have observed optimal performance in Chrome. The most notable difference across OS/browsers is likely to be in display handling for the PDF preview in the main panel: various user or browser-specific settings will determine the default zoom level.
|
|
|
|
## Citing shinyDAG
|
|
|
|
shinyDAG was developed by Jordan Creed, Garrick Aden-Buie and Travis Gerke.
|
|
|
|
Concept DOI: 10.5281/zenodo.1288712
|
|
|
|
v0.1.0 DOI: 10.5281/zenodo.1296477
|
|
|
|
v0.0.0 DOI: 10.5281/zenodo.1288713
|
|
|
|
## References
|
|
|
|
1. Richard Iannone (NA). DiagrammeR: Graph/Network Visualization. R package version 1.0.0.
|
|
[https://github.com/rich-iannone/DiagrammeR](https://github.com/rich-iannone/DiagrammeR).
|
|
1. Johannes Textor and Benito van der Zander (2016). dagitty: Graphical Analysis of Structural Causal
|
|
Models. R package version 0.2-2. [https://CRAN.R-project.org/package=dagitty](https://CRAN.R-project.org/package=dagitty).
|
|
1. Malcolm Barrett (2018). ggdag: Analyze and Create Elegant Directed Acyclic Graphs. R package
|
|
version 0.1.0. [https://CRAN.R-project.org/package=ggdag](https://CRAN.R-project.org/package=ggdag).
|
|
1. Csardi G, Nepusz T: The igraph software package for complex network research, InterJournal,
|
|
Complex Systems 1695. 2006. [http://igraph.org](http://igraph.org).
|
|
1. Hernan MA, Hernandez-Díaz S, Robins JM. A structural approach to selection bias. Epidemiology 2004;15:615-625.
|