baselinenowcast (version 0.2.0)

preprocess_negative_values: Preprocess negative values in the reporting triangle

Description

Takes in a reporting triangle and returns it with negative values of reporting handled by redistributing them to earlier delays (from longer delay to shorter). This is useful when dealing with reporting corrections that can result in negative incremental counts.

When negative values are detected, they are set to zero and the negative amount is subtracted from the count at the next earlier delay (moving from right to left in each row). This process continues until either the negative value is fully absorbed or the first delay is reached.

This code was adapted from code written (under an MIT license) by the Karlsruhe Institute of Technology RESPINOW German Hospitalization Nowcasting Hub. Modified from https://github.com/KITmetricslab/RESPINOW-Hub/blob/main/code/baseline/functions.R #nolint

Usage

preprocess_negative_values(reporting_triangle, validate = TRUE)

Value

A reporting_triangle object with negative values handled via redistribution to earlier delays.

Arguments

reporting_triangle

A reporting_triangle object.

validate

Logical. If TRUE (default), validates the object. Set to FALSE only when called from functions that already validated.

Details

Use this function when:

• Your data contains reporting corrections that result in negative counts

• You want to preserve the total count while handling negatives

• You need a delay distribution that sums to 1 or a CDF that is weakly increasing

Do not use this function when:

• Your data naturally has negative PMF entries (e.g., from differencing)

• You want to preserve the original structure including negatives

• You are working with corrections that should be reflected as negative probabilities

See Also

Delay distribution estimation functions estimate_delay()

Examples

# Using example dataset with negative values from corrections
# Preprocess to handle negatives
preprocessed <- preprocess_negative_values(example_downward_corr_rt)
preprocessed