cases: Distinguish between Cases Specified by Logical Conditions

In memisc: Management of Survey Data and Presentation of Analysis Results

Distinguish between Cases Specified by Logical Conditions

Description

cases allows to distinguish several cases defined logical conditions. It can be used to code these cases into a vector. The function can be considered as a multi-condition generalization of ifelse.

Usage

cases(...,check.xor=c("warn","stop","ignore"),
      .default=NA,.complete=FALSE,
      check.na=c("warn","stop","ignore"),
      na.rm=TRUE)

Arguments

...	A sequence of logical expressions or assignment expressions containing logical expressions as "right hand side".
check.xor	character (either "warn", "stop", or "ignore") or logical; if TRUE or equal to "stop" or "warn", cases checks whether the case conditions are mutually exclusive. If this is not satisfied and check.xor equals "warn" (the default), a warning is shown, otherwise an error exception is raised.
.default	a value to be used for unsatisfied conditions.
.complete	logical, if TRUE an additional factor level is created for the unsatisfied conditions.
check.na	character (either "warn", "stop", or "ignore") or logical; if TRUE or equal to "stop" or "warn", cases checks, whether any of the case conditions evaluates to NA. If that case, if check.na is TRUE or equals "stop" an error exception is raised, while if check.na equals "warn" (the default) a warning is shown.
na.rm	a logical value; how to handle NAs (if they do not already lead to an error exception). If FALSE if any of the conditions evaluates to NA, the corresponding value of the result vector is NA. If TRUE (the default), the resulting vector or factor is NA only for instances where all conditions result in NA.

Details

There are two distinct ways to use this function. Either the function can be used to construct a factor that represents several logical cases or it can be used to conditionally evaluate an expression in a manner similar to ifelse.

For the first use, the ... arguments have to be a series of logical expressions. cases then returns a factor with as many levels as logical expressions given as ... arguments. The resulting factor will attain its first level if the first condition is TRUE, otherwise it will attain its second level if the second condition is TRUE, etc. The levels will be named after the conditions or, if name tags are attached to the logical expressions, after the tags of the expressions. Not that the logical expressions all need to evaluate to logical vectors of the same length, otherwise an error condition is raised. If .complete is TRUE then an additional factor level is created for the conditions not satisfied for any of the cases.

For the second use, the ... arguments have to be a series of assignment expression of the type <expression> <- <logical expression> or <logical expression> -> <expression>. For cases in which the first logical expression is TRUE, the result of first expression that appears on the other side of the assignment operator become elements of the vector returned by cases, for cases in which the second logical expression is TRUE, the result of the second expression that appears on the other side of the assignment operator become elements of the vector returned by cases, etc. For cases that do not satisfy any of the given conditions the value of the .default argument is used. Note that the logical expressions also here all need to evaluate to logical vectors of the same length. The expressions on the other side of the assignment operator should also be either vectors of the same length and mode or should scalars of the same mode, otherwise unpredictable results may occur.

Value

If it is called with logical expressions as ... arguments, cases returns a factor, if it is called with assignment expressions the function returns a vector with the same mode as the results of the "assigned" expressions and with the same length as the logical conditions.

Examples

# Examples of the first kind of usage of the function
#
df <- data.frame(x = rnorm(n=20), y = rnorm(n=20))
df <- df[do.call(order,df),]
(df <- within(df,{
  x1=cases(x>0,x<=0)
  y1=cases(y>0,y<=0)
  z1=cases(
    "Condition 1"=x<0,
    "Condition 2"=y<0,# only applies if x >= 0
    "Condition 3"=TRUE
    )
  z2=cases(x<0,(x>=0 & y <0), (x>=0 & y >=0))
  }))
xtabs(~x1+y1,data=df)
dd <- with(df,
  try(cases(x<0,
            x>=0,
            x>1,
            check.xor=TRUE)# let's be fussy
            )
  )
dd <- with(df,
  try(cases(x<0,x>=0,x>1))
  )
genTable(range(x)~dd,data=df)

# An example of the second kind of usage of the function:
# A construction of a non-smooth function
#
fun <- function(x)
  cases(
    x==0      -> 1,
    abs(x)> 1 -> abs(x),
    abs(x)<=1 -> x^2
  )
x <- seq(from=-2,to=2,length=101)
plot(fun(x)~x)

# Demo of the new .default and .complete arguments
x <- seq(from=-2,to=2)
cases(a = x < -1,
      b = x > 1,
      .complete = TRUE)
cases(x < -1,
      x > 1,
      .complete = TRUE)
cases(1 <- x < -1,
      3 <- x > 1,
      .default = 2)

threshhold <- 5
d <- c(1:10, NaN)

d1 <- cases(
  d > threshhold -> 1,
  d <= threshhold -> 2
)

d2 <- cases(
  is.na(d) -> 0,
  d > threshhold -> 1,
  d <= threshhold -> 2
)

# Leads to missing values because some of the conditions result in missing
# even though they could be 'captured'
d3 <- cases(
  is.na(d) -> 0,
  d > threshhold -> 1,
  d <= threshhold -> 2,
  na.rm=FALSE
)

d4 <- cases(
  is.na(d) -> 0,
  d > threshhold +2 -> 1,
  d <= threshhold -> 2,
  na.rm=FALSE
)

cbind(d,d1,d2,d3,d4)

cases(
  d > threshhold,
  d <= threshhold
)

cases(
  is.na(d),
  d > threshhold,
  d <= threshhold
)

cases(
  d > threshhold,
  d <= threshhold,
  .complete=TRUE
)

cases(
  d > threshhold + 2,
  d <= threshhold,
  .complete=TRUE
)

memisc documentation built on Oct. 24, 2024, 5:07 p.m.