LGDtoolkit (version 0.2.0)

stepRPC: Stepwise (OLS & fractional logistic) regression based on risk profile concept

Description

stepRPC customized stepwise regression with p-value and trend check which additionally takes into account the order of supplied risk factors per group when selects a candidate for the final regression model. Trend check is performed comparing observed trend between target and analyzed risk factor and trend of the estimated coefficients. Note that procedure checks the column names of supplied db data frame therefore some renaming (replacement of special characters) is possible to happen. For details, please, check the help example.

Usage

stepRPC(
  start.model,
  risk.profile,
  p.value = 0.05,
  db,
  reg.type = "ols",
  check.start.model = TRUE,
  offset.vals = NULL
)

Value

The command stepRPC returns a list of four objects.

The first object (model), is the final model, an object of class inheriting from "glm".

The second object (steps), is the data frame with risk factors selected at each iteration.

The third object (warnings), is the data frame with warnings if any observed. The warnings refer to the following checks: if risk factor has more than 10 modalities or if any of the bins (groups) has less than 5% of observations.

The final, fourth, object dev.db returns the model development database.

Arguments

start.model

Formula class that represents the starting model. It can include some risk factors, but it can be defined only with intercept (y ~ 1 where y is target variable).

risk.profile

Data frame with defined risk profile. It has to contain the following columns: rf and group. Column group defines order of groups that will be tested first as a candidate for the regression model. Risk factors selected in each group are kept as a starting variables for the next group testing. Column rf contains all candidate risk factors supplied for testing.

p.value

Significance level of p-value of the estimated coefficients. For numerical risk factors this value is is directly compared to the p-value of the estimated coefficients, while for categorical risk factors multiple Wald test is employed and its value is used for comparison with selected threshold (p.value).

db

Modeling data with risk factors and target variable. All risk factors (apart from the risk factors from the starting model) should be categorized and as of character type.

reg.type

Regression type. Available options are: "ols" for OLS regression and "frac.logit" for fractional logistic regression. Default is "ols". For "frac.logit" option, target has to have all values between 0 and 1.

check.start.model

Logical (TRUE or FALSE), if risk factors from the starting model should checked for p-value and trend in stepwise process.

offset.vals

This can be used to specify an a priori known component to be included in the linear predictor during fitting. This should be NULL or a numeric vector of length equal to the number of cases. Default is NULL.

Examples

library(monobin)
library(LGDtoolkit)
data(lgd.ds.c)
num.rf <- sapply(lgd.ds.c, is.numeric)
num.rf <- names(num.rf)[!names(num.rf)%in%"lgd" & num.rf]
num.rf
for	(i in 1:length(num.rf)) {
num.rf.l <- num.rf[i]
lgd.ds.c[, num.rf.l] <- sts.bin(x = lgd.ds.c[, num.rf.l], y = lgd.ds.c[, "lgd"])[[2]]	
}
str(lgd.ds.c)
#define risk factor groups
set.seed(123)
rf.pg <- data.frame(rf = names(lgd.ds.c)[!names(lgd.ds.c)%in%"lgd"], 
		  group = sample(1:5, ncol(lgd.ds.c) - 1, rep = TRUE))
rf.pg <- rf.pg[order(rf.pg$group, rf.pg$r), ]
rf.pg
res <- LGDtoolkit::stepRPC(start.model = lgd ~ 1, 
	   risk.profile = rf.pg, 
	   p.value = 0.05, 
	   db = lgd.ds.c,
	   reg.type = "ols")
names(res)
summary(res$model)$coefficients
summary(res$model)$r.squared