02_PracticalDataConcerns_challenge_markdown.Rmd

---
title: 'PPAS Challenge: Practical data concerns'
output: 
  html_document:
    toc: True
  pdf_document:
    toc: True
---

## Background
In this challenge we use a [Kaggle dataset](https://www.kaggle.com/mazharkarimi/heart-disease-and-stroke-prevention/metadata) with data on the prevalence of cardiovascular disease and risk factors. We have created a synthetic, derivative dataset for the purposes of this challenge. The synthetic dataset contains 5 years of seriatim data on heart attack rates by state, year, sex, age, and race.

## Data license
The database license and content license that govern the original dataset can be found in a document in the PPAS GitHub repository. 

## Goals
Prepare data for modeling and validation.

## Load data and packages
```{r Load, warning = F, message = F}
library(dplyr)
library(car)
library(stringr)

heartattack <- readRDS("heartdiseasedataset_modified.RDS")
```

## Review data
```{r DataReview}
head(heartattack)
str(heartattack)
```

## Challenges
### 1) Deal with outlier and missing values

a) Find the number of missing values in each field.

```{r 1a_NumNA}

# YOUR CODE HERE ####

```

b) Does the missingness in any field correlate to the response variable or to other fields' values? Consider using the table() function for cross-tabulating two categorical variables.

```{r 1b_NACorrelation}

# YOUR CODE HERE ####

```

c) Impute the missing values.

```{r 1c_Imputation}

# YOUR CODE HERE ####

```

### 2) Derive a new variable for "geographic region" of USA to reduce dimensionality of that field.

```{r 2_DeriveRegion}

# YOUR CODE HERE ####

```

### 3) Partition off a 30% holdout subset

```{r 3_PartitionHoldout}

# YOUR CODE HERE ####

```

### 4) Fit logistic regression to estimate heart attack odds given region, year, age, sex, and race.

```{r 4_FitLogisticModel}

# YOUR CODE HERE ####

```


### 5) Test for multicollinearity between the predictor variables.

```{r 5_TestMulticollinearity}

# YOUR CODE HERE ####

```