cc_regression_analysis_code.Rmd

---
title: "Multiple Regression Analysis of Credit Card Data"
author: "Niko Seino"
date: "2023-03-13"
output: html_document
---
### Business Question:
1. What variables effectively contribute to predicting active cardholders'
credit card balances?
2. What credit card balance might a new active cardholder hold
depending on certain variables?

Install and load packages
```{r setup, include=FALSE}
library(tidyverse)
library(lm.beta)
library(olsrr)
```

Import and view data
```{r}
setwd("~/")
carddf <- read_csv("Credit.csv")
View(carddf)
```
Check for any missing values
```{r}
sum(is.na(carddf))
```
Convert categorical variables to factors
```{r}
carddf$Student <- factor(carddf$Student, levels = c(0, 1), labels = c("No","yes"))
carddf$Gender <- factor(carddf$Gender, levels = c(0, 1), labels = c("Male","Female"))
carddf$Married <- factor(carddf$Married, levels = c(0, 1), labels = c("No","yes"))
```


Generate summary statistics
```{r}
summary(carddf)
```
Partition the data into a training set and a validation set
(Since it is a small dataset, will divide 50-50)
```{r}
set.seed(42)
sample <- sample(c(TRUE, FALSE), nrow(carddf), replace=TRUE, prob=c(0.5,0.5))
traincard  <- carddf[sample, ]
validatecard <- carddf[!sample, ]
```

Create a correlation matrix for the quantitative variables in the training dataframe
```{r}
cor(traincard[c(-6, -7, -8)])
```
Conduct multiple regression analysis using the training dataframe with 'Balance'
as the outcome variable, and all others as predictor variables. View the summary. 
```{r}
card_MR <- lm(Balance ~ Income + Limit + Rating + Age + Education + Student + Gender + Married, traincard)
summary(card_MR)
```
Calculate Variance Inflation Factor for each predictor variable to assess 
multicollinearity
```{r}
vif(card_MR)
```
Conduct a multiple regression analysis with 'Balance' as the outcome variable and all other variables
except 'Limit' as predictor variables
```{r}
card_MR2 <- lm(Balance ~ Income + Rating + Age + Education + 
    Student + Gender + Married, data = traincard)
summary(card_MR2)
```
Create a residual plot using these results
```{r}
card_predict = predict(card_MR2)
card_resid = resid(card_MR2)
resid_df <- data.frame(card_predict, card_resid)

ggplot(resid_df, aes(x= card_predict, y = card_resid)) +
  geom_point() +
  labs(title = "Residual Plot", x = "Predicted Values", y = "Residuals")
```
Create a probability plot
```{r}
card_stdres <- rstandard(card_MR2)    #get standardized residuals
qqnorm(card_stdres, ylab = "Standardized residuals", xlab = "Normal scores")
```

Create a new regression analysis using the training dataframe with 'Balance' as
the outcome variable and only the statistically significant variables as predictors
```{r}
cardSS_MR <- lm(Balance ~ Income + Rating + Age + Student, traincard)
summary(cardSS_MR)
```
Run a standardized regression with standardized variables
```{r}
lm.beta(cardSS_MR)
```


Conduct a final multiple regression analysis with the validation dataframe with 'Balance' as 
the outcome variable and statistically significant variables as the predictors
```{r}
validcard_df <- lm(Balance ~ Income + Rating + Age + Student, validatecard)
summary(validcard_df)
```
Predict the credit card balances of the new cardholders in credic_card_prediction.csv
with 95% prediction intervals
```{r}
setwd("~/")
newcard <- read_csv("credit_card_prediction.csv")
newcard$Student <- factor(newcard$Student, levels = c(0, 1), labels = c("No","yes"))

predict(validcard_df, newcard, interval = "prediction", level = .95)
```