In GabyP/categorizaOrdinales: What the Package Does (One Line, Title Case)

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)
options(tibble.print_min = 5, tibble.print_max = 5)

options(warn = -1)

Overview

This is a package that groups variables (categorical and numerical) according to their likelihood of producing the event analyzed. It can be used as a tool to explore information for building a model to predict a binary target.

Install

library(devtools)
#devtools::install_github("gabriela-plantie/preparation4modeling", force=T, dependencies = F)
library(preparation4modeling)

Usage

Example table

set.seed(1)
x1 = rnorm(1000)
x2 = rnorm(1000)
x4='A'
x4=ifelse(x1>0.1,'B', x4)
x4=ifelse(x1>0.4,'C', x4 )
x4=ifelse(x1>0.6,'D', x4 )
x4=ifelse(x1>0.8,'E', x4 )
z = 1 + 3*x1
pr = 1/(1+exp(-z))
y = rbinom(1000,1,pr)
tbla = data.frame(y=y,x1=x1,x2=x2, x4=x4)
q_nas=100
x1[1:q_nas] = NA
x4[1:q_nas]=NA

Analyzing variables against target

numerical variable with tree

agrupa_ctree (tbla, target_name='y', variable_name='x1',flag_numerica=1, max_q_groups=10, algoritmo='chaid' )

categorical variable with tree

agrupa_ctree (tbla, target_name='y', variable_name='x4',flag_numerica=0, algoritmo='chaid' )

categorical variable with hipergeometric test

agrupa_nominal_filtra_small(tbla, target_name='y', variable_name='x4',limite=0.05, symbol_to_split='%#%', limite_grupo=100)

example table 2

x1 = rnorm(1000)
x2 = rnorm(1000)
x3= ifelse(as.factor(x2>0.5)==T, 'A', 'B')
x4= ifelse(as.factor(x2>0.7)==T, 'C', 'D')
z = 1 + 2 * x1 + 3 * x2
pr = 1/(1+exp(-z))
y = rbinom(1000,1,pr)
tbla = data.frame(y=y,x1=x1,x2=x2, x3=x3, x4=x4)

define level according to bad rate

tbla<-redefine_level_0( df_agrupada_y=tbla ,variables=c('x3',  'x4') ,nombre_target='y')

generating model and scorecard table

filtros_train= (tbla$random=runif(nrow(tbla)))<0.5
f=formula(y~x3+x4)
lr <- glm(f, tbla[ filtros_train, ], family = 'binomial')
tabla_estimadores(lr)

example table 3

x1 = rnorm(1000)
x2 = rnorm(1000)
z = 1 + 2 * x1 + 3 * x2
pr = 1/(1+exp(-z))
y = rbinom(1000,1,pr)
y1 = rbinom(1000,1,abs(pr-0.05))
tbla = data.frame(y=y,x1=x1,x2=x2, y1=y1)
f=formula(y~x1+x2)
lr <- glm(f, tbla, family = 'binomial')
tbla$prob<-predict(lr, tbla, type='response')

generating performance table

ventiles(tbla, targets=c('y', 'y1'), score_name = 'prob')

GabyP/categorizaOrdinales documentation built on Sept. 21, 2020, 1:42 p.m.