Stepwise covariate-parameter regressions

This function is largely superseded as it is accessed through the $step methods for PM_result and PM_cov objects. There is rarely a need to call it directly any longer.

Usage

PMstep(x, icen = "median", direction = "backward")

Arguments

x: A PMcov object which is the $data field of a PM_cov object
icen: A character vector to summarize covariate values. Default is "median", but can also be "mean".
direction: The direction for covariate elmination can be "backward" (default), "forward", or "both".

Value

A matrix with covariates in the rows and parameters in the columns. Values for the matrix are the multi-variate P-values. A value of NA indicates that the variable was not retained in the final model.

Details

It will perform stepwise linear regressions on a PM_cov object or to maintain backwards compatibility on a PMcov object made by makeCov. Every covariate in the model will be tested in a stepwise linear regression for their relationships to each parameter in the model. Bayesian posterior parameters and individual covariates are used.

Author