ReevesDogeUSD.Rmd

---
title: "ReevesDOGE_USD"
author: "Sam Reeves"
date: "4/11/2021"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```
We're going to showcase some features of various packages from the Tidyverse!

Here we have data concerning the DogeCoin/USD exchange rates according to Yahoo Finance from 2014 to roughly the present.  Using magrittr, dplyr, lubridate, and ggplot2, this should be a snap.  Let's get started.

### Load the data.

```{r}
rates <- read.csv("https://raw.githubusercontent.com/TheWerefriend/data607/master/tidyverseAssignment/DOGE-USD.csv")

colnames(rates)
```

### Check to see if close and adjusted close are the same.

```{r}
tests <- ifelse(rates$Close==rates$Adj.Close, "Identical", "Distinct")

"Distinct" %in% tests
```
This result makes sense because there are no possible corporate actions like stock splits which can cause the closing value in a day to be adjusted.

### Remove Adj.Close.

```{r message=FALSE}
library(magrittr)
library(dplyr)

rates <- rates %>%
  select(!Adj.Close)

colnames(rates)
```

From the dplyr package, we can use the ! operator to select the complement of a list of variables by name.

### Convert the Date column to a date, and everything else to numeric values.

```{r message=FALSE}
library(lubridate)

rates$Date <- ymd(rates$Date)
```

Now, all the other values are stored as factors... But, something strange happens when we try to convert them directly to numeric values:

```{r}
rates$Open[[1]]
as.numeric(rates$Open[[1]])
```
Apparently, 0.000293 is the 190th factor level in the column!  Since there are some null values in the data (days where major changes happened with the DOGE network, I assume), we must throw throw those observations out.

```{r warning=FALSE}
rates[, c(2:6)] <- rates %>%
  select(!Date) %>%
  sapply(as.character) %>%
  sapply(as.numeric)
```
```{r}
rates <- na.omit(rates)
anyNA(rates)
```

Transmute() works exactly the same as mutate(), except that it alters an existing variable instead of creating a new one.

### Graph the prices and volume of Doge.

```{r}
library(ggplot2)

a <- ggplot(rates, aes(x = rates$Date)) +
  geom_line(aes(y = rates$Open), color = "steelblue") +
  geom_line(aes(y = rates$High), color = "green") +
  geom_line(aes(y = rates$Low), color = "red") +
  geom_line(aes(y = rates$High), color = "purple")

b <- ggplot(rates, aes(x = rates$Date)) +
  geom_line(aes(y = rates$Volume), color = "black")

a + labs(x = "Time", y = "Price in USD")
b + labs(x = "Time", y = "Volume")
```

### TidyVerse Extend -- Richard Zheng (github: schoolkidrich)

```{r}
# remove outliers using group_by() and filter()
rates2 = rates %>%
  group_by(Close)%>%
  filter(between(Close, mean(Close) - (3 * sd(Close)), 
                     mean(Close) - (3 * sd(Close))))

# scatter plot with line of best fit
rates2 %>%
  ggplot(aes(x = Volume, y = Close))+geom_point()+geom_smooth(method = 'lm', formula = y~x)

# linear model predicting Close price by Volume 
m1 = lm(Close~log(Volume),rates2)
summary(m1)


# histogram of residuals show mostly normal distribution
m1 %>%
  resid()%>%
  hist()

# predicted values
rates$predict_close = predict(m1,rates)

# predicted close prices show much less volatility than what really happened (removing outliers caused my model to be innacurate)
rates%>%
  ggplot(aes(x=Date,y=predict_close))+geom_line(color = 'red')+geom_line(aes(x = Date,y=Close),color='blue')

# model 2 without removing outliers and using open price
m2 = lm(Close~Open+Volume,rates)
summary(m2)

# predictions by model
rates$predict_close2 = predict(m2)

# m2 seems to fit the data well 
rates%>% 
  ggplot(aes(x=Date))+geom_line(aes(y = Close), color='blue')+geom_line(aes(y = predict_close2), color='red')

```