Test Your Knowledge on Building Linear Regression Models!

Evaluated skills

Build Linear Regression Models

Description

In this quiz, we're going to build several models based on the bike-sharing dataset.

The bike -sharing dataset is available from the UCI repository.

The dataset has over 17k samples and 16 different variables. We are going to focus on the following five attributes:

Season: season (1:spring, 2:summer, 3:fall, 4:winter).
Temp: Normalized temperature in Celsius.
Hum: Normalized humidity.
Wind speed: Normalized wind speed.
Cnt: count of total rental bikes.

You can load the dataset with:

import pandas as pd
df = pd.read_csv('bike_sharing_day.csv')

Remove the non-essential columns with:

df = df[['season', 'temp','hum','windspeed','cnt']]

To do this quiz, you should first import the following packages:

import matplotlib.pyplot as plt
import seaborn as sns
import statsmodels.formula.api as smf
import statsmodels.api as sm
import numpy as np
import pandas as pd

Question 1
Let's find out which predictor is driving the usage of bikes (the cnt outcome variable). First, build the three univariate regression models:
1. cnt ~ temp
2. cnt ~ hum
3. cnt ~ wind speed
Looking at the R-squared metric, which variable explains the most the variability in the outcome cnt?
- Temp
- Hum
- Wind speed
Question 2
Look at the influence of each predictor for the different seasons.
The seasons are defined as:
```
seasons = {1:'spring', 2:'summer', 3:'fall', 4:'winter'}
```
You can limit the regression to a specific season, for instance, spring, with the following line:
```
res = smf.ols(formula, data = df[df.season == 1]).fit()
```
Looking at the R-squared for each season and univariate model:
- cnt ~ temp
- cnt ~ hum
- cnt ~ windspeed
Which if the following assertion below is true?

Don't hesitate to loop over the seasons and the predictors with the following code:
```
seasons = {1:'spring', 2:'summer', 3:'fall', 4:'winter'}

for season in range(1,5):
    print("--"* 20)
    print("season {}".format(seasons[season]))
    for variable in ['temp', 'hum', 'windspeed']:
        formula = "cnt ~ {}".format(variable)
        res = smf.ols(formula, data = df[df.season == season]).fit()
        print("- R^2 for {}: {:.2f}".format(variable, res.rsquared))
```
- Temp has the most influence in spring.
- Humidity is the most important factor in the fall.
- Wind speed always has very little influence on the usage.
- All of the above.
Question 3
When looking at the data over all four seasons, all p-values are well below 0.05, and the three predictors are relevant.
However, when selecting a specific season, some predictors are no longer significant.

Consider the p-values of the predictors in each univariate model for each season:
- cnt ~ temp
- cnt ~ hum
- cnt ~ windspeed
Which assertion is true?
- Humidity is never significant.
- Temperature is not significant in the fall.
- Wind speed is always a significant factor.
- None of the predictors are significant in winter.

Test Your Knowledge on Building Linear Regression Models!

Evaluated skills

Description

Question 1

Question 2

Question 3