R/reduceformula.R

Defines functions reduceformula

Documented in reduceformula

#
#  reduceformula.R
#
#  $Revision: 1.7 $   $Date: 2016/12/30 01:44:07 $
#
# delete variable from formula 
#
#......................................................
#

reduceformula <- function(fmla, deletevar, verbose=FALSE) {
  ## removes the variable `deletevar' from the formula `fmla'
  ## returns a simplified formula, or NULL if it can't simplify.
  stopifnot(inherits(fmla, "formula"))
  stopifnot(is.character(deletevar) && length(deletevar) == 1)
  if(!(deletevar %in% all.vars(as.expression(fmla)))) {
    if(verbose)
      message(paste("The formula does not involve", dQuote(deletevar),
                    "and is therefore unchanged"))
    return(fmla)
  }
  lhs <- if(length(fmla) < 3) NULL else fmla[[2]]
  ## create terms object
  tt <- attributes(terms(fmla))
         ##  formula.has.intercept <- (tt$intercept == 1)
  ## extract all variables appearing in the model
  vars <- as.list(tt$variables)[-1]
  nvars <- length(vars)
  varexprs <- lapply(vars, as.expression)
  varstrings <- sapply(varexprs, paste)
  ## identify any offsets
  offs <- tt$offset
  v.is.offset <- if(!is.null(offs)) (1:nvars) %in% offs else rep(FALSE, nvars)
  ## remove the response
  repo <- tt$response
  if(repo != 0) {
    vars <- vars[-repo]
    varstrings <- varstrings[-repo]
    varexprs <- varexprs[-repo]
    v.is.offset <- v.is.offset[-repo]
  }
  ## a term may be a variable name
           ##  v.is.name <- sapply(vars, is.name)
  ## a term may be an expression like sin(x), poly(x,y,degree=2)
  v.args <- lapply(varexprs, all.vars)
  matches.delete <- lapply(v.args, "==", deletevar)
  v.has.delete <- sapply(matches.delete, any)
  v.has.other <- !sapply(matches.delete, all)
  v.is.mixed <- v.has.delete & v.has.other
  ## we can't handle mixed terms like sin(x-d), poly(x,d)
  ## where d is to be deleted. Handling these would require
  ## knowledge about the functions sin and poly.
  if(any(v.is.mixed)) {
    nmixed <- sum(v.is.mixed)
    if(verbose)
      message(paste("Don't know how to reduce the",
              ngettext(nmixed, "term", "terms"),
              paste(dQuote(varstrings[v.is.mixed]), collapse=",")))
    return(NULL)
  }
  ## OK. We have identified all first order terms to be deleted.
  condemned.names <- varstrings[v.has.delete]
  ## Determine the terms of all orders that include these first order terms
  ## (1) terms with model coefficients
  fax <- tt$factors
  if(prod(dim(fax)) == 0)
    retained.terms <- character(0)
  else {
    ## Rows are first order terms 
    condemned.row <- rownames(fax) %in% condemned.names
    ## Columns are the terms of all orders
    allterms <- colnames(fax)
    ## Find all columns containing a 1 in a row that is to be deleted
    if(any(condemned.row)) {
      condemned.column <- matcolany(fax[condemned.row, , drop=FALSE] != 0)
      retained.terms <- allterms[!condemned.column]
    } else retained.terms <- allterms
  }
  ## (2) offsets if any
  if(any(v.is.offset))
    retained.terms <- c(retained.terms,
                        varstrings[v.is.offset & !v.has.delete])
  ## (3) intercept forced?
  if(length(retained.terms) == 0)
    retained.terms <- "1"
  
  ## OK. Cut-and-paste
  f <- paste(lhs, "~", paste(retained.terms, collapse=" + "))
  return(as.formula(f))
} 

Try the spatstat.core package in your browser

Any scripts or data that you put into this service are public.

spatstat.core documentation built on May 18, 2022, 9:05 a.m.