From fe31a790eeed2401ae659a9a070eece75847269d Mon Sep 17 00:00:00 2001 From: "Logan C. Brooks" Date: Wed, 4 Oct 2023 14:47:54 -0700 Subject: [PATCH] `document()` --- man/layer_cdc_flatline_quantiles.Rd | 17 +++++++++++++---- 1 file changed, 13 insertions(+), 4 deletions(-) diff --git a/man/layer_cdc_flatline_quantiles.Rd b/man/layer_cdc_flatline_quantiles.Rd index 71f414e25..22219ba7f 100644 --- a/man/layer_cdc_flatline_quantiles.Rd +++ b/man/layer_cdc_flatline_quantiles.Rd @@ -39,12 +39,21 @@ in linear interpolation on the X scale. This is achieved with \item{by_key}{A character vector of keys to group the residuals by before calculating quantiles. The default, \code{c()} performs no grouping.} -\item{symmetrize}{logical. If \code{TRUE} then interval will be symmetric.} +\item{symmetrize}{Logical. If \code{TRUE}, does two things: (i) forces the +"empirical" CDF of residuals to be symmetric by pretending that for every +actually-observed residual X we also observed another residual -X, and (ii) +at each ahead, forces the median simulated value to be equal to the point +prediction by adding or subtracting the same amount to every simulated +value. Adjustments in (ii) take place before propagating forward and +simulating the next ahead. This forces any 1-ahead predictive intervals to +be symmetric about the point prediction, and encourages larger aheads to be +more symmetric.} \item{nonneg}{Logical. Force all predictive intervals be non-negative. -Because non-negativity is forced \emph{before} propagating forward, this -has slightly different behaviour than would occur if using -\code{\link[=layer_threshold]{layer_threshold()}}.} +Because non-negativity is forced \emph{before} propagating forward, this has +slightly different behaviour than would occur if using \code{\link[=layer_threshold]{layer_threshold()}}. +Thresholding at each ahead takes place after any shifting from +\code{symmetrize}.} \item{id}{a random id string} }