gainsChart: Generate a gains chart

In dnegrey/miscTools: R package with miscellaneous tools

Description

gainsChart produces a gains chart table, evaluating y as the dependent variable and yhat as the vector of predicted values. When yhat is a combined model score, use y2 to supply the other dependent variable. Users may specify the number of bins (i.e. 10 for deciles) that are desired.

Usage

gainsChart(y, yhat, y2 = NULL, numBins = 10)

Arguments

y	logical, integer or numeric vector (dependent variable)
yhat	numeric vector (predicted values of y)
y2	logical, integer or numeric vector (other dependent variable for combined models)
numBins	integer value >= 2; number of desired bins

Value

A data frame with class "mt_gainsChart" containing the following columns:

• yhatBin: integer specifying the bin number (i.e. decile)

• minYhat: minimum value of yhat

• maxYhat: maximum value of yhat

• Freq: number of observations (relative)

• cumFreq: number of observations (cumulative)

• sumY: sum of y values (relative)

• sumYB (combined only): sum of y (binary) values (relative)

• avgY: average of y values (relative)

• avgYB (combined only): average of y (binary) values (relative)

• Lift: lift of avgY (relative)

• LiftYB (combined only): lift of avgYB (relative)

• sumYC (combined only): sum of y (continuous) values (relative)

• avgYC (combined only): average of y (continuous) values (relative)

• LiftYC (combined only): lift of avgYC (relative)

• perFreqYC (combined only): average (per Freq) of y (continuous) values (relative)

• LiftYC2 (combined only): lift of perFreqYC (relative)

• cumSumY: sum of y values (cumulative)

• cumSumYC (combined only): sum of y (continuous) values (cumulative)

• cumAvgY: average of y values (cumulative)

• cumLift: lift of avgY (cumulative)

• cumPctY: percentage of y values (cumulative)

• cumPctYC (combined only): percentage of y (continuous) values (cumulative)

• nonY (binary only): sum of non-y values (relative)

• cumNonY (binary only): sum of non-y values (cumulative)

• cumPctNonY (binary only): percentage of non-y values (cumulative)

• cumPctRandom (continuous & combined): uniform random percentage (cumulative)

• KS: K-S statistic

See Also

glm, lm

Examples

# pull in sample scored data frame
x <- modelSampleScored
# filter to validation subset
x <- x[x$ValidateFlag, ]
# gains chart binary
gainsChart(x$TargetFlag, x$pTargetFlag)
# gains chart continuous
gainsChart(x[x$TargetFlag, "TargetValue"], x[x$TargetFlag, "pTargetValue"])
# gains chart combined
gainsChart(x$TargetFlag, x$pTargetFlag*x$pTargetValue, x$TargetValue)

dnegrey/miscTools documentation built on May 3, 2019, 2:57 p.m.