all_the_R_you_need_tidy.Rmd

---
title: "All the R you need for STAT 220 - tidy"
author: "Professor McNamara"
date: "Fall 2020"
output: html_document
---

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

## Packages

We use three packages in this course:

```{r, message=FALSE}
library(Lock5Data)
library(tidyverse)
library(infer)
```

## Reading in data

```{r, message=FALSE}
GSS <- read_csv("data/GSS_clean.csv")
```

## Plotting


```{r}
# one categorical
ggplot(GSS) + geom_bar(aes(x = marital_status))

# two categorical
ggplot(GSS) + geom_bar(aes(x = marital_status, fill = born_in_us))
ggplot(GSS) + geom_bar(aes(x = marital_status, fill = born_in_us), position = "dodge")

# one numeric
ggplot(GSS) + geom_histogram(aes(x = highest_year_of_school_completed), binwidth = 2)
ggplot(GSS) + geom_boxplot(aes(x = 1, y = highest_year_of_school_completed))

# one numeric, one categorical
ggplot(GSS) + geom_boxplot(aes(x = marital_status, y = highest_year_of_school_completed))

# two numeric
ggplot(GSS) + geom_point(aes(
  x = highest_year_of_school_completed,
  y = highest_year_school_completed_spouse
))
```

## Summary statistics


```{r}
# one categorical
GSS |>
  group_by(marital_status) |>
  summarize(n = n()) |>
  mutate(prop = n / sum(n))

# two categorical
GSS |>
  group_by(marital_status, general_happiness) |>
  summarize(n = n()) |>
  mutate(prop = n / sum(n))

# one numeric
GSS |>
  summarize(
    mean = mean(highest_year_of_school_completed, na.rm = TRUE),
    median = median(highest_year_of_school_completed, na.rm = TRUE)
  )

GSS |>
  drop_na(highest_year_of_school_completed) |>
  summarize(
    mean = mean(highest_year_of_school_completed),
    median = median(highest_year_of_school_completed)
  )

GSS |>
  drop_na(highest_year_of_school_completed) |>
  summarize(sd(highest_year_of_school_completed))

GSS |>
  drop_na(highest_year_of_school_completed) |>
  summarize(
    range = max(highest_year_of_school_completed) - min(highest_year_of_school_completed),
    IQR = IQR(highest_year_of_school_completed)
  ) # range and IQR

GSS |>
  drop_na(highest_year_of_school_completed) |>
  summarize(
    min = min(highest_year_of_school_completed),
    lower_hinge = quantile(highest_year_of_school_completed, .25),
    median = median(highest_year_of_school_completed),
    upper_hinge = quantile(highest_year_of_school_completed, .75),
    max = max(highest_year_of_school_completed)
  ) # five number summary

GSS |>
  pull(highest_year_of_school_completed) |>
  fivenum() # five number summary

GSS |>
  summarize(correlation = cor(
    x = highest_year_of_school_completed,
    y = highest_year_school_completed_spouse,
    use = "complete.obs"
  )) # correlation
```

## Working with data

```{r}
# fix NA issue
GSS <- GSS |>
  drop_na(highest_year_of_school_completed)

# turn character to numeric
GSS <- GSS |>
  mutate(Age = parse_number(age_of_respondent))

# filter observations
GSS |>
  filter(Age > 91)
```

## Modeling/inference

```{r}
# linear model
m1 <- lm(highest_year_of_school_completed ~ highest_year_school_completed_spouse,
  data = GSS
)
m1 |>
  summary() # model summary

m1 |> predict(
  newdata = data.frame(highest_year_school_completed_spouse = 9),
  interval = "confidence"
)
m1 |> predict(
  newdata = data.frame(highest_year_school_completed_spouse = 9),
  interval = "prediction"
)

# bootstrapping
bootstrap_sample <- GSS |>
  specify(response = Age) |>
  generate(reps = 1, type = "bootstrap")

boot <- GSS |>
  specify(response = Age) |>
  generate(reps = 1000, type = "bootstrap") |>
  calculate(stat = "mean")

# one proportion
GSS |>
  prop_test(
    response = self_emp_or_works_for_somebody,
    explanatory = NULL, 
    alternative = "greater", p = 0.1
  )

# two proportions
GSS |>
  prop_test(response = should_marijuana_be_made_legal, 
            explanatory = self_emp_or_works_for_somebody, 
            order = c("Self-employed", "Someone else"))

# one mean
GSS |>
  drop_na(born_in_us) |>
  t_test(response = highest_year_of_school_completed, 
         explanatory = NULL,
         mu = 12, alternative = "greater")

# two means
GSS |>
  drop_na(born_in_us) |>
  t_test(
    response = highest_year_of_school_completed,
    explanatory = born_in_us, order = c("No", "Yes")
  )

# ANOVA
a1 <- aov(Age ~ marital_status, data = GSS)
a1 |> 
  summary()
a1 |> 
  TukeyHSD(conf.level = 0.90) # differences

# Chi-square
GSS |>
  chisq_test(response = marital_status, explanatory = general_happiness) # independence

GSS |>
  chisq_test(response = general_happiness, p = c(0.14, 0.56, 0.30)) # GoF
```

## Probability

```{r}
pnorm(1.96, lower.tail = FALSE) # normal distribution function
qnorm(0.25, mean = 100, sd = 5) # normal quantile function
```