2021_06_08_great_lakes_fish.Rmd

---
title: "TidyTemplate"
date: 2021-06-08
output: html_output
---

# TidyTuesday

Join the R4DS Online Learning Community in the weekly #TidyTuesday event!
Every week we post a raw dataset, a chart or article related to that dataset, and ask you to explore the data.
While the dataset will be “tamed”, it will not always be tidy! As such you might need to apply various R for Data Science techniques to wrangle the data into a true tidy format.
The goal of TidyTuesday is to apply your R skills, get feedback, explore other’s work, and connect with the greater #RStats community!
As such we encourage everyone of all skills to participate!

```{r setup, include=FALSE}

knitr::opts_chunk$set(echo = TRUE)

library(tidyverse)
library(tidytuesdayR)
library(scales)
theme_set(theme_light())
```

# Load the weekly Data

Dowload the weekly data and make available in the `tt` object.

```{r Load}

tt <- tt_load("2021-06-08")

fishing <- tt$fishing %>%
  filter(values >= 0) %>%
  mutate(species = str_replace(str_to_title(species), "([^s])s$", "\\1"))

summarize_fishing <- function(tbl) {
  tbl %>%
    summarize(n_obs = n(),
              total_production = sum(values, na.rm = TRUE)) %>%
    arrange(desc(total_production))
}

fishing %>%
  group_by(species) %>%
  summarize_fishing()

fishing %>%
  ggplot(aes(values + 1)) +
  geom_histogram() +
  scale_x_log10()

fishing %>%
  count(lake)

fishing %>%
  count(species, sort = TRUE)

fishing %>%
  group_by(year, lake, species) %>%
  summarize(total_values = sum(values, na.rm = TRUE),
            first_grand_total = min(grand_total, na.rm = TRUE),
            n_grand_total = n_distinct(grand_total, na.rm = TRUE)) %>%
  ungroup() %>%
  count(n_grand_total)

fishing %>%
  filter(year == 2000, lake == "Erie", species == "Carp")

fishing %>%
  group_by(decade = 10 * year %/% 10) %>%
  summarize_fishing() %>%
  ggplot(aes(decade, total_production)) +
  geom_col() +
  scale_y_continuous(labels = comma_format()) +
  labs(x = "Decade",
       y = "Total production per decade (thousands of pounds)")

fishing %>%
  group_by(species) %>%
  summarize_fishing()

fishing %>%
  mutate(species = fct_lump(species, 15, w = values),
         species = fct_reorder(species, values, sum, .desc = TRUE)) %>%
  group_by(decade = 10 * year %/% 10,
           species) %>%
  summarize_fishing() %>%
  ggplot(aes(decade, total_production, fill = species)) +
  geom_area() +
  scale_y_continuous(labels = comma_format()) +
  facet_wrap(~ species) +
  theme(legend.position = "none") +
  labs(x = "Decade",
       y = "Total production per decade (thousands of pounds)",
       title = "Production of 9 common fish in Great Lakes over time")

fishing %>%
  group_by(species, year) %>%
  summarize_fishing() %>%
  summarize(total_production = sum(total_production),
            peak_year = year[which.max(total_production)]) %>%
  arrange(desc(total_production)) %>%
  head(25) %>%
  mutate(species = fct_reorder(species, peak_year)) %>%
  ggplot(aes(peak_year, species)) +
  geom_point(aes(size = total_production)) +
  scale_size_continuous(labels = comma_format()) +
  labs(x = "Year of peak production",
       y = "",
       size = "All-time production")
```

```{r}
fishing %>%
  mutate(lake = fct_reorder(lake, values, sum, .desc = TRUE)) %>%
  group_by(decade = 10 * year %/% 10, lake) %>%
  summarize_fishing() %>%
  ggplot(aes(decade, total_production, fill = lake)) +
  geom_area() +
  scale_y_continuous(labels = comma_format()) +
  facet_wrap(~ lake) +
  theme(legend.position = "none") +
  labs(x = "Decade",
       y = "Total production per decade (thousands of pounds)",
       title = "Production across the Great Lakes over time")
```

```{r}
by_lake_species <- fishing %>%
  filter(lake != "Saint Clair") %>%
  mutate(species = fct_lump(species, 20, w = values),
         species = fct_reorder(species, values, sum),
         lake = fct_reorder(lake, values, sum, .desc = TRUE)) %>%
  group_by(lake, species) %>%
  summarize_fishing()

by_lake_species %>%
  ggplot(aes(lake, species, fill = total_production)) +
  geom_tile() +
  scale_fill_gradient2(low = "white", high = "darkblue",
                       labels = comma_format()) +
  expand_limits(fill = 0) +
  theme(panel.grid = element_blank()) +
  labs(x = "Lake",
       y = "Species",
       fill = "All-time Production")

by_lake_species %>%
  group_by(lake) %>%
  mutate(pct = total_production / sum(total_production)) %>%
  ggplot(aes(lake, species, fill = pct)) +
  geom_tile() +
  scale_fill_gradient2(low = "white", high = "darkblue",
                       labels = percent_format(accuracy = 1)) +
  expand_limits(fill = 0) +
  theme(panel.grid = element_blank()) +
  labs(x = "Lake",
       y = "Species",
       fill = "% of lake's production")
```

```{r}
stocked <- tt$stocked %>%
  janitor::clean_names()

stocked %>%
  count(site, sort = TRUE)

stocked %>%
  count(st_site, sort = TRUE)

stocked %>%
  count(species, sort = TRUE)

stocked %>%
  ggplot(aes(year)) +
  geom_histogram(binwidth = 1)

stocked %>%
  filter(!is.na(grid))

stocked %>%
  count(stat_dist, sort = TRUE)

stocked %>%
  count(stage)

stocked %>%
  ggplot(aes(length)) +
  geom_histogram() +
  scale_x_log10()

stocked %>%
  group_by(species) %>%
  summarize(median_length = median(length, na.rm = TRUE))
```

Forecasting

```{r}
fishing %>%
  filter(species == "Yellow Perch") %>%
  group_by(year, species) %>%
  summarize_fishing() %>%
  ggplot(aes(year, total_production)) +
  geom_line()

by_year_species <- fishing %>%
  mutate(year = as.Date(paste0(year, "-01-01"))) %>%
  group_by(year, species) %>%
  summarize_fishing() %>%
  ungroup() %>%
  select(-n_obs)

library(timetk)
library(forecast)
library(sweep)
library(lubridate)

yellow_perch <- by_year_species %>%
  filter(species == "Yellow Perch")

time_series <- yellow_perch %>%
  tk_ts(start = min(year(.$year)), freq = 1)

ets_mod <- time_series %>%
  ets()

ets_mod %>%
  sw_tidy()

ets_mod %>%
  sw_augment() %>%
  ggplot(aes(index, .actual)) +
  geom_line() +
  geom_line(aes(y = .fitted), color = "blue") +
  labs(x = "Year",
       y = "Total production of Yellow Perch")

fcast <- ets_mod %>%
    forecast(h = 10) 

fcast_holt <- holt(time_series)
fcast_ses <- ses(time_series)

autoplot(time_series) +
  autolayer(fcast_holt, series="Holt", PI = FALSE) +
  autolayer(fcast_ses, series="SES", PI = FALSE)
```

```{r}
time_series <- by_year_species %>%
  filter(species == "Cisco") %>%
  select(-species) %>%
  tk_ts(start = min(year(.$year)), freq = 1)

forecasts <- by_year_species %>%
  mutate(species = fct_lump(species, 8)) %>%
  group_by(species, year) %>%
  summarize(total_production = sum(total_production), .groups = "drop") %>%
  nest(data = c(-species)) %>%
  mutate(time_series = map(data, ~ tk_ts(., start = min(year(.$year)), freq = 1))) %>%
  mutate(holt = map(time_series, holt, h = 20),
         ses = map(time_series, ses, h = 20))

forecasts %>%
  mutate(forecast_sweep = map(holt, sw_sweep)) %>%
  unnest(forecast_sweep) %>%
  ggplot(aes(index, total_production)) +
  geom_line() +
  geom_ribbon(aes(ymin = lo.80, ymax = hi.80), alpha = .2) +
  facet_wrap(~ species)

fcast_holt <- holt(time_series, h = 50)
fcast_holt_damped <- holt(time_series, damped = TRUE, h = 50)
fcast_ses <- ses(time_series, h = 50)

autoplot(time_series) +
  autolayer(fcast_holt, series="Holt", PI = FALSE) +
  autolayer(fcast_holt_damped, series="Holt (Damped)", PI = FALSE) +
  autolayer(fcast_ses, series="SES", PI = FALSE)
```