monobin (version 0.2.4)

ndr.bin: Four-stage monotonic binning procedure including regression with nested dummies

Description

ndr.bin implements extension of three-stage monotonic binning procedure (iso.bin) with step of regression with nested dummies as fourth stage. The first stage is isotonic regression used to achieve the monotonicity. The next two stages are possible corrections for minimum percentage of observations and target rate, while the last regression stage is used to identify statistically significant cut points.

Usage

ndr.bin(
  x,
  y,
  sc = c(NA, NaN, Inf, -Inf),
  sc.method = "together",
  y.type = NA,
  min.pct.obs = 0.05,
  min.avg.rate = 0.01,
  p.val = 0.05,
  force.trend = NA
)

Value

The command ndr.bin generates a list of two objects. The first object, data frame summary.tbl

presents a summary table of final binning, while x.trans is a vector of discretized values. In case of single unique value for x or y of complete cases (cases different than special cases), it will return data frame with info.

Arguments

x

Numeric vector to be binned.

y

Numeric target vector (binary or continuous).

sc

Numeric vector with special case elements. Default values are c(NA, NaN, Inf, -Inf). Recommendation is to keep the default values always and add new ones if needed. Otherwise, if these values exist in x and are not defined in the sc vector, function will report the error.

sc.method

Define how special cases will be treated, all together or separately. Possible values are "together", "separately".

y.type

Type of y, possible options are "bina" (binary) and "cont" (continuous). If default value is passed, then algorithm will identify if y is 0/1 or continuous variable.

min.pct.obs

Minimum percentage of observations per bin. Default is 0.05 or 30 observations.

min.avg.rate

Minimum y average rate. Default is 0.05 or 30 observations.

p.val

Threshold for p-value of regression coefficients. Default is 0.05. For a binary target binary logistic regression is estimated, whereas for a continuous target, linear regression is used.

force.trend

If the expected trend should be forced. Possible values: "i" for increasing trend (y increases with increase of x), "d" for decreasing trend (y decreases with decrease of x). Default value is NA. If the default value is passed, then trend will be identified based on the sign of the Spearman correlation coefficient between x and y on complete cases.

See Also

iso.bin for three-stage monotonic binning procedure.

Examples

suppressMessages(library(monobin))
data(gcd)
age.bin <- ndr.bin(x = gcd$age, y = gcd$qual)
age.bin[[1]]
table(age.bin[[2]])
#linear regression example
amount.bin <- ndr.bin(x = gcd$amount, y = gcd$qual, y.type = "cont", p.val = 0.05)
#create nested dummies
db.reg <- gcd[, c("qual", "amount")]
db.reg$amount.bin <- amount.bin[[2]]
amt.s <- db.reg %>% 
      group_by(amount.bin) %>% 
      summarise(qual.mean = mean(qual),
		    amt.min = min(amount))
mins <- amt.s$amt.min
for (i in 2:length(mins)) {
	 level.l <- mins[i]
 nd <- ifelse(db.reg$amount < level.l, 0, 1)
 db.reg <- cbind.data.frame(db.reg, nd)
 names(db.reg)[ncol(db.reg)] <- paste0("dv_", i)
 }
reg.f <- paste0("qual ~ dv_2 + dv_3")
lrm <- lm(as.formula(reg.f), data = db.reg)
lr.coef <- data.frame(summary(lrm)$coefficients)
lr.coef
cumsum(lr.coef$Estimate)
#check
as.data.frame(amt.s)
diff(amt.s$qual.mean)