Classe Prediction Assignment-1.Rmd

---
title: "Classe Prediction Assignment"
author: "Sankarshan Acharya"
date: "July 16, 2016"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```


## The Model

In this prediction project, I did an unusual method for predicting the classe that each of the particitpants were in. From my research, I looked up what the concept of ordered and nominal variable research was.By definition, classe is a nominal variable. Therefore I used the specialized regression techniques associated with nominal variables. 

In R, there are packages that are meant to handle nominal variable regression. This package is called MASS. In MASS, there is a function called polr. This function does for ordered variables what glm does for numeric variables. What you do is pick a model type and see what kind of model you want to to fit. In this case, I used probit (there were other options to choose from). 

The independent variables that I used were the respective total acceleration from the belt, arm, forearm, and the dumbbell. 


```{r, echo=FALSE, include=FALSE}
require(MASS)
require(Hmisc)
require(reshape2)
```


```{r, echo = FALSE}
pml_training <- read.csv("~/pml-training.csv")
m <- polr(classe ~ total_accel_belt + total_accel_arm + total_accel_forearm + total_accel_dumbbell, data = pml_training, Hess=TRUE)
summary(m)
```

I also combined this in conjunction with a heurisitic model that I created from the below plots. This model looked for specific values from each of the acceleration graphs. By finding these values, I would be able to immediately tell what class the test subject is in. 

## Acceleration vs. Classes

The plots describe the statistics of acceleration vs. classe:

```{r, echo=FALSE}
pml_training <- read.csv("~/pml-training.csv")
boxplot(total_accel_belt ~ classe,data = pml_training, main="Total Belt Acceleration", xlab="Exercise Specification", ylab="Total Belt Acceleration")
boxplot(total_accel_dumbbell ~ classe, data = pml_training, main="Total Dumbbell Acceleration", xlab="Exercise Specification", ylab="Total Dumbbell Acceleration")
boxplot(total_accel_arm ~ classe,data = pml_training, main="Total Acceleration of the Upper Arm", xlab="Exercise Specification", ylab="Total Upper Arm Acceleration")
boxplot(total_accel_forearm ~ classe,data = pml_training, main="Total Acceleration of the Foreman", xlab="Exercise Specification", ylab="Total Forearm Acceleration")
```

Note that the `echo = FALSE` parameter was added to the code chunk to prevent printing of the R code that generated the plot.