createNAvars: Make NAs available for regression

In regr0: Building regression models

Description

For numeric variables, introduces a new indicator (binary) variable for missing values and assigns a fixed, non-missing value for them in the (renamed) original variable
For factors, introduces a new level for missing values and assigns it to all missing values.

Usage

  createNAvars(data, vars=NULL, na.prop=0.1, na.label=".NA.",
    na.values = NULL, name.suffix=c(".X",".NA"), append=TRUE, ...)
  factorNA(data, na.label = ".NA.",na.prop=0,  ...)
  xNA(data, na.values = NULL, na.prop=0.1, name.suffix=c(".X",".NA"))

Arguments

data	a data.frame of quantitative variables. A vector will be converted to a one column data.frame
vars	variables (column names or indexes) to be treated. It is usually not needed to specify vars since only those variables are treated that contain missing values.
na.prop	limit for the proportion of NAs. Only those variables are treated that contain a higher proportion of missings.
na.label	for factors, label to be used for the new level of the factor
na.values	for numeric variables, value used in the .X variable for the NAs
name.suffix	suffix to be used for the new quantitative variables and the new indicator variables
append	logical: should the modified any newly created variables be returned with the unmodified variables? If TRUE, factors will be replaced by the modified version and the created variables based on continuous variables will be appended.
...	other arguments, to be passed to the function factor

Value

data.frame, containing,

• for each numeric variable Vx, a variable Vx.X and a variable Vx.NA (if the default name.suffix is used),

• for each factor, the factor consisting of the original values, but with NAs replaced by a level '.NA.' (unless specified otherwise).

If append is true, the data.frame will be the same as the input data, except that factors containing NAs will have been modified and the variables created for numeric variables containing NAs will be appended. Note that no variables will be appended for those that did not contain any NAs.

Author(s)

Werner A. Stahel

See Also

factorNA

Examples

data(d.blast)
dd <- d.blast
dd$location[c(1,3,5)] <- NA
rr <- createNAvars(dd, na.prop=0)

xNA(c(3,NA,NA,6,5,9,1,11))

df <- data.frame(a=c(3,NA,NA,6,5,9,1,11), b=c(1,2,1,0,3,4,2,0))
( dfn <- xNA(df) )
attr(dfn, "xNA.values")

factorNA(c("a","c",NA,"a",NA), na.label="missing")

regr0 documentation built on May 2, 2019, 4:52 p.m.