Nowcasting nomenclature

Throughout this package, we will refer to different nowcasting object lists, matrices, and vectors using the following table. In this table, “point” refers to a point estimate. When not indicated, we are referring to a probabilistic draw from an observation model.

Data Structure	Observations Only	Mixed (Obs + Predictions)	Pure Predictions Only
Matrix Format	reporting_matrix (complete with observations) reporting_triangle (contains NAs for missing observations)	point_nowcast_matrix nowcast_matrix	point_pred_matrix pred_matrix (contains NAs in elements where there were observations)
List Format (multiple matrices)	reporting_matrices (complete) reporting_triangles (contains NAs for missing observations)	nowcast_matrices	pred_matrices
Vector Format (summed across delays)	observed_cases	point_nowcast nowcast	point_pred pred
DataFrame Format	-	nowcast_df	pred_df

For example, we refer to the matrix with imputed point estimates for all \(t+d>t^*\) as a point nowcast matrix, a matrix with a complete set of observations for all elements as a reporting matrix, and a matrix with only the predictions as a point prediction matrix.

In package documentation, we will generally use a reporting triangle to refer to a diagonal reporting triangle, which is the special case of a reporting triangle where the reporting delays and reference times are indexed on the same scale. An example of a diagonal reporting triangle is one with daily reporting of daily data. The package also supports what we refer to as a ragged reporting triangle, an example of which would be daily reporting of data referenced only once per week.

We will use the following to abbreviations to shorten names in the code when necessary:

Long Name	Code Abbreviation
truncated	trunc
retrospective	retro
observed	obs
predicted	pred
matrix	mat
point	pt
error	err