---
title: "Precept 10"
author: "Emily Nelson"
date: "April 13, 2016"
output: html_document
---

```{r global_options, include=FALSE}
knitr::opts_chunk$set(message=FALSE, cache=TRUE, warning=FALSE)
```


Agenda:

- Davis data -- 2 sample t-test and least squares regression example from class
- Linear Regression -- Intuitive Approach
- ANOVA and interpretation
- Do



```{r}
library(dplyr)
library(ggplot2)
library(reshape2)

#install.packages("car")
library("car")
data("Davis", package="car")

Davis = Davis %>%
  filter(height > 100)

```


#Two Sample T-Test and Least Squares Regression

A least squares regression using a factor variable is equivalent to performing a two-sample t-test.

```{r}
ggplot(data = Davis, aes(x=height, y=weight, color = sex)) +
  geom_point() +
  theme_bw() +
  scale_color_manual(values = c("red", "blue"))

davis_model <- lm(weight ~ sex, data=Davis)
summary(davis_model)

t.test(Davis[Davis$sex=="F",]$weight, Davis[Davis$sex=="M",]$weight,
       var.equal=TRUE)
```

Explain these values in **words**.

#Intuitive Interpretation of Multiple Least Squares

```{r}
model1 <- lm(height ~ weight, data=Davis)
summary(model1)

davis_fitted = Davis
davis_fitted$fitted = model1$fitted.values
  
ggplot(data = Davis, aes(x=weight, y=height, color=sex)) +
  geom_point() +
  geom_line(data = davis_fitted, aes(x=weight, y=fitted),
            size=2, color="black") +
  scale_color_manual(values = c("red", "blue")) +
  theme_bw()
```

```{r}
model2 <- lm(height ~ weight + sex, data=Davis)
summary(model2)

davis_fitted = Davis
davis_fitted$fitted = model2$fitted.values
  
ggplot(data = Davis, aes(x=weight, y=height, color=sex)) +
  geom_point() +
  geom_line(data = davis_fitted, aes(x=weight, y=fitted),
            size=2) +
  scale_color_manual(values = c("red", "blue")) +
  theme_bw()
```

```{r}
model3 <- lm(height ~ weight + sex + weight*sex, data=Davis)
summary(model3)

davis_fitted = Davis
davis_fitted$fitted = model3$fitted.values
  
ggplot(data = Davis, aes(x=weight, y=height, color=sex)) +
  geom_point() +
  geom_line(data = davis_fitted, aes(x=weight, y=fitted),
            size=2) +
  scale_color_manual(values = c("red", "blue")) +
  theme_bw()
```

```{r}
Davis_sq = Davis
Davis_sq$weight2 = Davis_sq$weight * Davis_sq$weight

model4 <- lm(height ~ weight + sex + weight2, data=Davis_sq)
summary(model4)

davis_fitted = Davis_sq
davis_fitted$fitted = model4$fitted.values
  
ggplot(data = Davis, aes(x=weight, y=height, color=sex)) +
  geom_point() +
  geom_line(data = davis_fitted, aes(x=weight, y=fitted),
            size=2) +
  scale_color_manual(values = c("red", "blue")) +
  theme_bw()
```

```{r}
model5 <- lm(height ~ weight + sex + weight2 + weight*sex, data=Davis_sq)
summary(model5)

davis_fitted = Davis_sq
davis_fitted$fitted = model5$fitted.values
  
ggplot(data = Davis, aes(x=weight, y=height, color=sex)) +
  geom_point() +
  geom_line(data = davis_fitted, aes(x=weight, y=fitted),
            size=2) +
  scale_color_manual(values = c("red", "blue")) +
  theme_bw()
```

What is happening as I add more and more terms? Can I just go on adding terms forever?

#ANOVA

```{r}
anova(model3)

anova(model5)
```

What does this tell us?

#Do or Do Not, There Is No Try

How do I efficiently fit multiple models?

```{r}
library(broom)

fit_model <- function(t) {
  m = lm(height ~ weight, t)
  return(tidy(m))
}

Davis %>%
  group_by(sex) %>%
  do(fit_model(.))

```