P8130_Final_Project_EDA_.Rmd

---
title: "EDA_Claire"
output: github_document
date: "2023-12-04"
---

```{r}
library(ggplot2)
library(corrplot)
library(dplyr)
library(tidyverse)
library(MASS)
```

#### Load Dataset

```{r}
df = read_csv("./data/Project_2_data.csv") |>
  janitor::clean_names() 

summary(df)
```

#### Correlation Analysis

```{r}
df$status = as.numeric(factor(df$status))
cols_num = which(sapply(df, is.numeric)) 
correlations = cor(df[cols_num])
correlations
```

From the results, we can see that `regional_node_examined` and `reginol_node_positive` have strong relationship with the correlation of 1.

```{r}
corrplot(correlations, order = "hclust", tl.cex = 1, addrect = 6)
```

From the plot, we can see that `reginol_node_positive` and `tumore_size` are top factors.

#### Univariate Analysis

```{r}
prop.table(table(df$status))
```

From the table, we can see that there are more than 84% of patients will be alive and about 15% patients will be dead.

```{r}
hist(df$survival_months, probability = T, main = 'original')

# Log Transformation
hist(log(df$survival_months), probability = T, main = 'log_transformed')
```

Left skewed distribution.

#### Bivariate analysis

```{r}
# Numerical Variables
par(mfrow = c(2,2))
boxplot(age ~ status, df)
boxplot(tumor_size ~ status, df)
boxplot(regional_node_examined ~ status, df)
boxplot(reginol_node_positive ~ status, df)
```

```{r}
# Categorical Variables
df_race = df |> 
  count(status, race)
ggplot(df_race, aes(fill = race, y = n, x = status)) + geom_bar(position = 'dodge', stat = 'identity') 
```

```{r}
df_marital = df |>
  count(status, marital_status)
ggplot(df_marital, aes(fill = marital_status, y = n, x = status)) + geom_bar(position = 'dodge', stat = 'identity') 
```

```{r}
df_tstage = df |>
  count(status, t_stage)
ggplot(df_tstage, aes(fill = t_stage, y = n, x = status)) + geom_bar(position = 'dodge', stat = 'identity') 
```

    
```{r}
df_nstage = df |>
  count(status, n_stage)
ggplot(df_nstage, aes(fill = n_stage, y = n, x= status)) + geom_bar(position = 'dodge', stat = 'identity') 
```