baselinenowcast

This is the first stable release providing core functionality for producing probabilistic or point nowcasts from a dataframe with incident case counts indexed by reference dates and report dates. It supports doing so across multiple strata, either through creating reporting_triangle objects for each strata and nowcasting independently, or nowcasting directly from a dataframe containing multiple strata. The reporting_triangle object allows users to specify the unit of the delay, and provides helpful print and summary S3 methods, as well as functionality to truncate to a specific delay or a percentile of the observed delays. A paper validating and evaluating the performance of the methods is available at <https://wellcomeopenresearch.org/articles/10-614/v1>.

Package

• fill_triangle() has been removed. Use estimate_and_apply_delay() instead, which now supports optional delay estimation via the delay_pmf parameter (#334).
• fill_triangles() has been removed. Use estimate_and_apply_delays() instead (#334).
• Rename truncate_triangle() to truncate_to_row() and truncate_triangles() to truncate_to_rows() to clarify that these functions truncate by row count and distinguish them from other truncation utilities such as truncate_to_quantile() and truncate_to_delay() (#336).
• Use the as_reporting_triangle and baselinenowcast workflow in the vignette which walks through a nowcasting example applied to syndromic surveillance data in the U.S.
• Add preprocess parameter to estimate_delay() to control handling of negative values in reporting triangles. Set to preprocess_negative_values by default to redistribute negative values to earlier delays, or set to NULL to preserve negative PMF entries. This allows the method to work with reporting corrections that result in net downward adjustments at specific delays (#278).
• Export preprocess_negative_values() function to allow users to manually handle negative values in reporting triangles by redistributing them to earlier delays (#278).
• Add a baselinenowcast.data.frame() method which ingests a data.frame with one or more strata to nowcast and returns a baselinenowcast_df object.
• Add converters between reporting_triangle and ChainLadder triangle formats via as_ChainLadder_triangle() and as_reporting_triangle.triangle(), enabling use of ChainLadder's mature claims reserving methods alongside baselinenowcast's nowcasting functionality.
• Add a baselinenowcast.reporting_triangle() method which ingests a reporting_triangle object and produces a baselinenowcast_df object which is a data.frame with nowcasts and associated metadata.
• Add a as_reporting_triangle() S3 method which ingests a data.frame or matrix and returns a reporting_triangle object, which will be used as an input to the eventual baselinenowcast() function.