---
title: "Models2"
author: "Eric A. Suess"
date: "2/22/2021"
output: html_document
---

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

## Model Google's stock price.

```{r}
library(pacman)
p_load(tidyverse, fpp3)
```


Re-index based on trading days

```{r}
google_stock <- gafa_stock %>%
  filter(Symbol == "GOOG", year(Date) >= 2015) %>%
  mutate(day = row_number()) %>%
  update_tsibble(index = day, regular = TRUE)

google_stock
```

Filter the year of interest

```{r}
google_2015 <- google_stock %>% filter(year(Date) == 2015)
google_2015 
```

Fit the models

```{r}
google_fit <- google_2015 %>%
  model(
    Mean = MEAN(Close),
    `Naïve` = NAIVE(Close),
    Drift = NAIVE(Close ~ drift())
  )
google_fit 
```

Produce forecasts for the trading days in January 2016

```{r}
google_jan_2016 <- google_stock %>%
  filter(yearmonth(Date) == yearmonth("2016 Jan"))

google_fc <- google_fit %>% 
  forecast(new_data = google_jan_2016)
google_fc
```

Plot the forecasts

```{r}
google_fc %>%
  autoplot(google_2015, level = NULL) +
  autolayer(google_jan_2016, Close, color = "black") +
  labs(x = "Day", y = "Closing Price (US$)",
       title = "Google stock prices (Jan 2015 - Jan 2016)") +
  guides(colour = guide_legend(title = "Forecast"))
```

```{r}
augment(google_fit)
```

## Residual diagnostics

A good forecasting method will yield innovation residuals with the following properties:

#. The innovation residuals are uncorrelated. If there are correlations between innovation residuals, then there is information left in the residuals which should be used in computing forecasts.
#. The innovation residuals have zero mean. If they have a mean other than zero, then the forecasts are biased.
#. The innovation residuals have constant variance.
#. The innovation residuals are normally distributed.


```{r}
autoplot(google_2015, Close) +
  labs(x = "Day", y = "Closing Price (US$)",
       title = "Google Stock in 2015")
```

Fit the Naive model (jus t the mean) and augment the dataset with the residuals and innovations.

```{r}
aug <- google_2015 %>%
  model(NAIVE(Close)) %>%
  augment()

autoplot(aug, .innov) +
  labs(x = "Day", y = "Residual",
       title = "Residuals from naïve method")
```

Normal?

```{r}
aug %>%
  ggplot(aes(x = .innov)) +
  geom_histogram() +
  labs(title = "Histogram of residuals")
```

```{r}
aug %>%
  ACF(.innov) %>%
  autoplot() +
  labs(title = "ACF of residuals")

```

```{r}
google_2015 %>%
  model(NAIVE(Close)) %>%
  gg_tsresiduals()
```

# Portmanteau test

From a French word describing a suitcase or coat rack carrying several items of clothing.

Test if there is at least one lagged autocorrelation is different from zero.

Box-Pierce

```{r}
aug %>% features(.innov, box_pierce, lag = 10, dof = 0)
```

Ljung-Box

```{r}
aug %>% features(.innov, ljung_box, lag = 10, dof = 0)
```