examine_driver_Ycat: Main driver analysis when Y is a categorical quantity

In regclass: Tools for an Introductory Class in Regression and Modeling

Main driver analysis when Y is a categorical quantity

Description

This function provides a "main driver analysis" on the association between a categorical y variable and the "driver" x. A visualization (mosaic plot) of the strength of the relationship is provided as well as numerical output to help quantify the variation of y across possible values of the "driver" x.

Usage

examine_driver_Ycat(formula,data,sort=TRUE,inside=TRUE,equal=TRUE)

Arguments

formula	A standard R formula written as y=="Yes"~x, where y is the variable of interest and "Yes" is the level of interest (you need to pick one of the levels of y to be the level of interest) and x is the driver.
data	An argument giving the name of the data frame that contains x and y.
sort	TRUE (default) or FALSE. Indicates whether to sort the avearge value of y from largest to smallest.
equal	If FALSE, the bar widths in the mosaic plot are proportional to the frequency of the corresponding level. If TRUE, the bar widths are all equal (useful if there are many levels or some are extremely rare).
inside	If FALSE, labels in the mosaic plot are beneath the bars. If TRUE, labels are placed inside the bars and rotated (useful if the levels have long names)

Details

Main driver analysis is a cornerstone of business analytics where we identify and quantify the key factors (drivers) that most strongly influence a business outcome or performance metric.

This function handles the case when y (the outcome variable) is categorical and you want to analyze the chance that an entity has a specific value of y (the level of interest). See examine_driver_Ynumeric when y is numeric.

This function works best if x is a categorical variable (with multiple examples of each level of x), since the probability that y equals the level of interest is estimated for each unique value of x.

A mosaic plot (see mosaic and its associated arguments) is presented to visualize the relationship between y and the driver.

A table giving the estimated probability that y has the level of interest for each value of x is provided. A "connecting letters report" shows which levels have statistically significant differences in the probability that y has the level of interest (if ANY letters are in common between two values of x, there is not a statistically significant difference in the probability that y has the level of interest between those two values of x; if ALL letters are different, the difference is statistically significant).

The function also provides a "Driver Score" (a value between 0 and 1; larger driver scores indicate stronger associations between the chance that y has the level of interest and x). This driver is the R-squared of a simple linear regression predicting 1 (y has the level of interest) or 0 (y does not have the level of interest) from x, and is at best treated as a relative score indicating the strength of the relationship (the value itself does not hold any practical significance).

Author(s)

Adam Petrie

References

Introduction to Regression and Modeling

See Also

examine_driver_Ynumeric,mosaic

Examples

	#No statistically significant differences in levels
  data(CUSTLOYALTY)
	examine_driver_Ycat(Married=="Single"~Income,data=CUSTLOYALTY)
	
	#Some statistically significant differences in levels
  data(EX6.CLICK)
	examine_driver_Ycat(Click=="Yes"~SiteID,data=EX6.CLICK)
	examine_driver_Ycat(Click=="Yes"~DeviceModel,data=EX6.CLICK)

	 

regclass documentation built on June 8, 2025, 12:40 p.m.