---
title: "Homework #10: Density Estimation" 
author: "**Your Name Here**"
format: ds6030hw-html
---

```{r config, include=FALSE}
# Set global configurations and settings here
knitr::opts_chunk$set()                 # set global chunk options
ggplot2::theme_set(ggplot2::theme_bw()) # set ggplot2 theme
```


# Required R packages and Directories {.unnumbered .unlisted}

```{r packages, message=FALSE, warning=FALSE}
data_dir = 'https://mdporter.github.io/teaching/data' # data directory
library(ks)        # functions for KDE
library(tidyverse) # functions for data manipulation   
```



# Problem 1 Geographic Profiling

Geographic profiling, a method developed in criminology, can be used to estimate the [home location (roost) of animals](https://www.sciencedirect.com/science/article/pii/S0022519305004157) based on a collection of sightings. The approach requires an estimate of the distribution the animal will travel from their roost to forage for food. 

A sample of $283$ distances that pipistrelle bats traveled (in meters) from their roost can be found at: 

- **Bat Data**: <`r file.path(data_dir, 'geo_profile.csv')`>

One probability model for the distance these bats will travel is:
\begin{align*}
f(x; \theta) = \frac{x}{\theta} \exp \left( - \frac{x^2}{2 \theta} \right)
\end{align*}
where the parameter $\theta > 0$ controls how far they are willing to travel. 


## a. Derive a closed-form expression for the MLE for $\theta$ (i.e., show the math). 

::: {.callout-note title="Solution"}
Add solution here.
:::


## b. Estimate $\theta$ for the bat data using MLE? 

Calculate using the solution to part a, or use computational methods.

::: {.callout-note title="Solution"}
Add solution here.
:::

## c. Plot the estimated density

Using the MLE value of $\theta$ from part b, calculate the estimated density at a set of evaluation points between 0 and 8 meters. Plot the estimated density.

- The x-axis should be distance and y-axis should be density (pdf). 

::: {.callout-note title="Solution"}
Add solution here.
:::

## d. Estimate the density using KDE. 

Report the bandwidth you selected and produce a plot of the estimated density. 

::: {.callout-note title="Solution"}
Add solution here.
:::

## e. Which model do you prefer, the parametric or KDE? Why?

::: {.callout-note title="Solution"}
Add solution here.
:::

# Problem 2: Interstate Crash Density

Interstate 64 (I-64) is a major east-west road that passes just south of Charlottesville. Where and when are the most dangerous places/times to be on I-64? The crash data (link below) gives the mile marker and fractional time-of-week for crashes that occurred on I-64 between mile marker 87 and 136 in 2016. The time-of-week data takes a numeric value of *\<dow\>.\<hour/24\>*, where the dow starts at 0 for Sunday (6 for Sat) and the decimal gives the time of day information. Thus `time=0.0417` corresponds to Sun at 1am and `time=6.5` corresponds to Sat at noon. 

- **Crash Data**: <`r file.path(data_dir, 'crashes16.csv')`>


## a. Crash Data

Extract the crashes and make a scatter plot with mile marker on x-axis and time on y-axis.

::: {.callout-note title="Solution"}
Add solution here.
:::

## b. Use KDE to estimate the *mile marker* density. 

- Report the bandwidth. 
- Plot the density estimate.

::: {.callout-note title="Solution"}
Add solution here.
:::

## c. Use KDE to estimate the temporal *time-of-week* density. 

- Report the bandwidth. 
- Plot the density estimate. 

::: {.callout-note title="Solution"}
Add solution here.
:::

## d. Use KDE to estimate the bivariate mile-time density. 

- Report the bandwidth parameters.
- Plot the bivariate density estimate. 

::: {.callout-note title="Solution"}
Add solution here.
:::

## e. Crash Hotspot

Based on the estimated density, approximate the most dangerous place and time to drive on this stretch of road. Identify the mile marker and time-of-week pair (within a few miles and hours).

::: {.callout-note title="Solution"}
Add solution here.
:::