Skip to contents

Return strongly-correlated parameter pairs

get_cor_pairs.Rd

Computes the parameter correlation matrix from the (inverse) Hessian and returns a data.frame of pairs whose absolute correlation exceeds threshold. Pairs are sorted by descending absolute correlation. If no pairs meet the criterion the function returns an informative string instead of a data.frame.

Usage

get_cor_pairs(
  obj,
  h = NULL,
  threshold = 0.9,
  include = c("core", "all"),
  digits = 3
)

Arguments

obj

An RTMB AD object from MakeADFun after optimisation.

h

Optional pre-computed Hessian matrix (square, same dimension as length(obj$par)). When NULL (default) the Hessian is obtained via obj$he() if available, otherwise optimHess.

threshold

Absolute correlation threshold. Only pairs with |cor| > threshold are returned. Default 0.9.

include

Which parameters to consider: "core" (default) excludes rdev_y parameters; "all" includes everything in obj$par.

digits

Number of significant digits applied to the correlation column. Default 3. Use NULL to suppress rounding.

Value

A data.frame with columns par1, par2, correlation (sorted by descending |correlation|), or a character string when no pairs exceed threshold.