This repository has collected different work from the Temporal Group from the DQ-CDM Code-a-thon Nov 6-8, 2015.
The Anomaly Detective provides an interactive app to explore temporal anomalies on top of DQ-CDM data.
To run the Shiny app, you will need to install R, then from within R you can install the required R packages as follows:
install.packages(c("shiny", "ggplot2", "data.table", "lubridate", "magrittr",
"gridExtra", "ggfortify", "forecast", "changepoint", "strucchange"),
dependencies=TRUE)Then you can run the app from within Rstudio by clicking Run App after opening either
ui.R or server.R.
For a selected condition (concept_id) and data source, the Shiny app presents results of different analyses in 4 tabs:
The first tab on the app overviews the data and calculates a "typical" annual pattern. It highlights data points that statistically deviate from the "typical" pattern, defined as falling outside the error bounds given by the user-selected multiple of the standard error (SE) estimate.
We compute a "typical" year by averaging over each month over all of the data. The bounds are a multiple of the SE, which the user can select. For the condition below, we see a seasonality effect, with higher prevalence in the winter months.
The second tab returns a table of high prevalence of the selected condition.
The third tab produces plots from a set of time series analyses.
Picture goes below...
The fourth tab plots prevelance overtime of the selected concept_id by database (for every database available).
