---
title: "1st homework assignment"
output: pdf_document
highlight: zenburn
---

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

## Task 1 - cleaning data (2 points)

Work with the **customer_behaviour** dataset. 

```{r}
load("customer_behaviour.RData")
```

The dataset has 4 columns, each row represents an individual customer: *money_spent* describes the average amount of money customer spends during one visit, *age* is self-explanatory, *web_visits* describes how many times a month customer checks out the shop website, *mail_ads* describes how many advertisement emails the customer gets monthly, *shop_visits* described how many times the customer visits a shop in person a month. Explore each variable and **delete** any rows which have mistakes in them. **Do not fix the mistakes, delete whole rows.**

| number of rows in the cleaned dataset | 
| -------------- | 
| `insert number` | 


## Task 2 - descriptive statistics (3 points)

Work with the cleaned dataset from the previous month **customer_behaviour2**.

```{r}
load("customer_behaviour2.RData")
```

Firstly, create a new variable called `big` where each value equals either 1 (if the person spent more money than 5000 USD), or 0 (if he spent less or equal):

```{r}
# insert your code here
```

Plot two boxplots of the variable *money_spent* into one figure: the first one for observations with the value of *big* equal to 0, the second one for observations with the value of *big* equal to 1. Then create a histogram for the variable *money_spent* together with its kernel density estimation.

```{r, echo = FALSE, out.width="49%"}
# insert your code here

```

Finally, compute following numerical characteristics of the variable *age*:

| mean | median | $1^{st}$ quartile | $3^{rd}$ quartile | interquartile range | variance |
| -------------- | --------------- | --------------- | --------------- | --------------------- | -------------- |
| `insert` | `insert` | `insert` | `insert` | `insert` | `insert` |

Choose one appropriate measure of location and one appropriate measure of variability for the *money_spent* variable. Input the name of the measure into the following table. Briefly explain why you chose these measures.

|measure of location | measure of variability |
|-----------------|-------------------|
| `insert name`| `insert name`|
|--------------------|---------------|
| `insert explanation` | `insert explanation` |

## Task 3 - correlation (2 points)

Compute the correlation matrix of the data from the previous task (excluding the *money_spent* and *big* variables) and the sum of all its diagonal elements. Explain the result of the sum:

```{r, include = F}
# compute the correlation matrix here:

# compute the sum of all diagonal elements here:

```

| Sum of diagonal elements | Explanation |
|---------------|--------------------------------|
| `number` | `text` |


Compute and **interpret** correlation coeficients between following variables:

| Variables | Results | Interpretation |
|-----------------|-----------|--------------------------------------|
| `example` | 0 | The correlation is zero, which means... |
| `money_spent`, `age` | `insert`| `insert` |
| `money_spent`, `web_visits` | `insert`| `insert` |

Interpretation in the form "correlation coefficient is 0.8 which means the correlation is high" will not be accepted.


## Task 4 - PCA (3 points)

Use PCA on the dataset from the previous task (**customer_behaviour2**, excluding variables *money_spent* and *big*). Use as little components as possible to capture at least 80 % of data variance. 

| Number of components used | 
|-----------------|
| `insert` | 

State which variable has the most influence on each component. 

| - | Component 1 | Component 2 | Component 3 | Component 4 | 
|-----------------|-----------------|-----------------|-----------------| -----------------|
most impactful variable | `insert` | `insert` | `insert` | `insert` | 
 

Create a scatter plot of data points using the first two components. Plot the points in different colours depending on the value of the *big* variable. What is your evaluation of the final plot? Can you decipher from the plot which variable(s) seems best at separating *big* shoppers from the customers who spend less?

```{r, include = T}

```

| Scatter plot evaluation | Which variable(s) best separates heavy spenders |
|-----------------|-----------------|
| `insert text explaining what can you see in the scatter plot` | `insert name of the variable(s) and explain why you came to such conclusion` |