Visualize traffic safety data and analyze the relationship between collisions and smartphone usage over time

**Inspect the traffic safety dataset**

After running the first cell to load all necessary libraries, we need to load our dataset. Using pandas, load the dataset `traffic.csv` and save it as `traffic`. Inspect the first few rows. 

**Inspect and format data types**

The `traffic` data frame contains three columns: `Date`, `Crashes_per_100k`, and `Season`. In order to plot the `Crashes_per_100k` column as a time series, we need to make sure that the `Date` column is in date format. Inspect the data types in the data frame, convert the `Date` column to date format, and inspect the data types a second time.

**Visualize traffic safety data**

To get a sense of trends that may exist in the data, use seaborn's `sns.lineplot()` function to create a line plot of the `traffic` data with `Date` on the x-axis and `Crashes_per_100k` on the y-axis. 

**Visualize seasonal rates**

Since we saw a fair amount of variance in the number of collisions occurring throughout the year, we might hypothesize that the number of collisions increases or decreases during different seasons. We can visually explore this with a box plot. 

Use `sns.boxplot()` with crash rate on the x-axis and season on the y-axis. Remove the anomalous 2020 data by adjusting the `data` parameter to `traffic[traffic.Date.dt.year != 2020]`.

**Inspect the smartphone use dataset**

The dataset `crashes_smartphones.csv` contains smartphone data from Pew Research Center matched to normalized crash rates from the `traffic` data frame for the years 2011 to 2019.

<details>
    <summary style="display:list-item;"><b>Toggle for an overview of the variables in this dataset.</b></summary>

* `Month_Year`: a shortened date with only the month and year of the survey
* `Crashes_per_100k`: the normalized crash rate matching the month and year of the smartphone usage survey
* `Season`: Winter, Spring, Summer, or Fall
* `Smartphone_Survey_Date`: the actual date the smartphone usage survey was conducted
* `Smartphone_usage`: the percent of survey participants that owned and used a smartphone

</details>

Load the dataset as `smartphones` and inspect the first few rows.

**Format date data type**

Similar to the  `traffic` data frame, the `smartphones` data frame has a date column that is not properly formatted. Convert the `Smartphone_Survey_Date` column to the date data type using the `pd.to_datetime()` function and then inspect the data types in the data frame.

**Visualize smartphone use data**

Now let's take a look at smartphone use over time. Create a line plot of the `smartphones` data with `Smartphone_Survey_Date` on the x-axis and `Smartphone_usage` on the y-axis.

**Visualize crash rate by smartphone use** 

A scatter plot with smartphone usage on one axis and crash rates on the other axis will give us an idea of whether there is a relationship between these two variables. 

Create a scatter plot with a regression line using seaborn's `sns.regplot()` with `Smartphone_usage` on the x-axis and `Crashes_per_100k` on the y-axis.  

**Check the correlation coefficient**

To test whether the correlation between `Smartphone_usage` and `Crashes_per_100k` is statistically significant, we can calculate the Pearson's _r_ correlation coefficient and the associated _p_-value. 

Use `corr, p = pearsonr(column1, column2)` on the `Smartphone_usage` and `Crashes_per_100k` columns in the `smartphones` dataframe. Then use the provided code to print `corr` and `p` to see the results.

**Run a linear regression**

We can use a linear regression to predict crash rates based on smart phone usage. Let's regress crash rates on smartphone usage. Then we can predict the crash rate in 2020 and see if it matches the actual crash rate in 2020!

We have provided the code to convert the variables to NumPy arrays that will work with the modeling function. The `Smartphone_usage` array is saved as `X`, and the `Crashes_per_100k` array is saved as `y`.

Initiate the model by saving `LinearRegression()` to the variable `lm`. Then fit the model and run the regression with `.fit()`.

**Print and interpret regression coefficients**

Let's see the values our model produced. Print the coefficients from our `lm` model. Then think about which parts of the regression line equation these values represent.

**Make a prediction**

Let's assume smartphone usage was the same for 2020 as it was for 2019. This is a reasonable asssumption since the increase in smartphone usage that we observed in our plot started to plateau at the end of the time series. Let's use this approximation and our regression model to predict the crash rate in 2020.

From our model output, the regression line equation is `Crashes_per_100k = 120.6637 + (0.6610 * Smartphone_usage)`. Run the provided code to view the smartphone usage rate for 2019. Then substitute this value into the equation, using Python as a calculator to predict the crash rate for 2020.

**Compare to the actual rate**

How good was our prediction? Get the actual crash rate for February of 2020 from the `traffic` data frame using `pd.to_datetime("2020-02-01")` as the value for `Date`.

**Visualize the prediction**

Let's plot our regression plot again, but let's add two new points on top:

* The predicted 2020 crash rate
* The actual 2020 crash rate

Code has been provided for the original regression plot and a legend title. 

Add a scatter plot layer to add the 2020 predicted and actual crash rates that both used the 2019 smartphone usage rate. Use different colors and marker shapes for the predicted and actual 2020 crash rates.