Lab3_Demo.Rmd

---
title: "Lab 3 Demo"
author: "Mateo Robbins"
date: "2023-01-24"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
library(tidyverse)
library(rsample)
library(skimr)
library(glmnet)
```

## Data Wrangling and Exploration
```{r data}
#load and inspect the data
dat <- AmesHousing::make_ames()

```

##Train a model
```{r intial_split}
# Data splitting with {rsample} 
set.seed(123) #set a seed for reproducibility
split <- 

ames_train <-  
ames_test  <- 

```

```{r model_data}
#Create training feature matrices using model.matrix() (auto encoding of categorical variables)

X <- 

# transform y with log() transformation
Y <- 

```

```{r glmnet}
#fit a ridge model, passing X,Y,alpha to glmnet()
ridge <- 

#plot() the glmnet model object

plot()  
```

```{r}
# lambdas applied to penalty parameter.  Examine the first few



# small lambda results in large coefficients

# what about for small coefficients?

  
```
How much improvement to our loss function as lambda changes?

##Tuning
```{r cv.glmnet}
# Apply CV ridge regression to Ames data.  Same arguments as before to glmnet()
ridge <- 

# Apply CV lasso regression to Ames data
lasso <- 
  
# plot results
par(mfrow = c(1, 2))
plot(ridge, main = "Ridge penalty\n\n")
plot(lasso, main = "Lasso penalty\n\n")
```

10-fold CV MSE for a ridge and lasso model. What's the "rule of 1 standard deviation"?

In both models we see a slight improvement in the MSE as our penalty log(λ) gets larger, suggesting that a regular OLS model likely overfits the training data. But as we constrain it further (i.e., continue to increase the penalty), our MSE starts to increase. 

Let's examine the important parameter values apparent in the plots.
```{r}
# Ridge model
# minimum MSE

# lambda for this min MSE

# 1-SE rule

# lambda for this MSE

# Lasso model

# minimum MSE

# lambda for this min MSE

# 1-SE rule

# lambda for this MSE

# No. of coef | 1-SE MSE
```

```{r}
# Ridge model
ridge_min 

# Lasso model
lasso_min


par(mfrow = c(1, 2))
# plot ridge model
plot(ridge_min, xvar = "lambda", main = "Ridge penalty\n\n")
abline(v = log(ridge$lambda.min), col = "red", lty = "dashed")
abline(v = log(ridge$lambda.1se), col = "blue", lty = "dashed")

# plot lasso model
plot(lasso_min, xvar = "lambda", main = "Lasso penalty\n\n")
abline(v = log(lasso$lambda.min), col = "red", lty = "dashed")
abline(v = log(lasso$lambda.1se), col = "blue", lty = "dashed")
```