MT5763 Coursework 1 - Housekeeping Solution

Submission
• This coursework is about creating a succinct and reproducible report with R Markdown, which is version controlled using Git, and hosted in a private GitHub repository. The report must contain a link to the private GitHub repository where it is hosted.
Marking guidance
• Code - readable, logical, reproducible, tidy and appropriately commented.
• Report - succinct and well-presented, with a consistent narrative throughout.
• Version control - use of Git / GitHub, exhibiting frequent and well-documented commits.
• Data wrangling - use of the pipe operator (%>%) for enhanced readability.
• Data visualisation - well-annotated graphs and a clear description of the insights gained.
• Statistical modelling - appropriate analysis and interpretation.
Assignment
Version control
• You only need to version control your *.Rmd file, nothing else.
• Make sure you commit changes often and include a succinct commit message that clearly describes what you changed and why. You will not be penalised for committing changes to fix mistakes in your code - this is one of the reasons why version control is used!
• Before submitting your coursework invite me as a collaborator to your repo so that I’m able to access it. Instructions can be found here. My username is jjvalletta.
• Include a link to your GitHub repo at the start of your R Markdown report.
Data description
There are two datasets you’ll be working on (both available to download from Moodle):
• BikeSeoul.csv
• BikeWashingtonDC.csv
Rental bike sharing systems have been introduced in many cities worldwide to provide an accessible and sustainable mode of transport. These datasets contain the number of bikes rented at each hour in Seoul, South Korea (BikeSeoul.csv) and Washington, D.C., USA (BikeWashingtonDC.csv), together with corresponding meteorological and holiday data.
We will use these two cities as examples to explore the relationships between bike usage, weather, time of day and holidays. Understanding these relationships is important to eventually build appropriate statistical models to predict bike demand at various times of the year. These predictions can then be used, for example, to schedule bike maintenance.
The datasets contain the following variables:
• BikeSeoul.csv
– Rented Bike count - Number of bikes rented in that hour
– Hour - Hour of the day
– Temperature - Air temperature in degree Celsius
– Humidity - As a %
– Windspeed - In m/s
– Visibility - In 10m units (i.e. visibility = 2000, means a 20km visibility)
– Dew point temperature - In degree Celsius
– Solar radiation - In MJ/m2
– Rainfall - In mm
– Snowfall - In cm
– Seasons - Winter, Spring, Summer, Autumn
– Holiday - Holiday / No holiday
– Functional Day - Yes / No bike count data collected
• BikeWashingtonDC.csv
– instant - Unique record index
– dteday - Day / Month / Year
– season - Season (1: Winter, 2: Spring, 3: Summer, 4: Autumn)
– yr - Year (0: 2011, 1:2012)
– mnth - Month
– hr - Hour
– holiday - 0: no holiday, 1: holiday
– weekday - Day of the week
– workingday - 0: holiday / weekend, 1: otherwise
– weathersit - Weather condition
1. clear, few clouds, partly cloudy
2. mist & cloudy, mist & broken clouds, mist & few clouds, mist
3. light snow, light rain & thunderstorm & scattered clouds, light rain & scattered clouds
4. Heavy rain & ice pellets & thunderstorm & mist, snow & fog
– temp : Normalised air temperature in degree Celsius. The values are computed via tmaxt−t−mintmin where tmin = −8◦C and tmax = +39◦C
– atemp: Normalised feeling temperature in degree Celsius. The values are computed via tmaxt−t−mintmin, where tmin = −16◦C and tmax = +50◦C
– hum: Normalised humidity. The values are divided by 100
– windspeed: Normalised wind speed. The values are divided by 67km/h (max)
– casual: Number of bikes rented by casual users
– registered: Number of bikes rented by registered users
– cnt: Total number of bikes rented in that hour (i.e. casual + registered)
Data wrangling
After reading the data in, first step is to clean it for downstream analysis. In particular, perform the following operations:
BikeSeoul.csv
• Remove the following columns: visibility, dew point temperature, solar radiation, rainfall and snowfall .
• Filter out observations for which no bike count data was collected, then remove the functioning day column as it is no longer required.
• Where necessary, change the name of the columns to the following names (you will do the same for the Washington data to have a consistent set of variable names across both datasets):
– Count - Number of bikes rented in that hour
– Hour - Hour of the day
– Temperature - Air temperature in degree Celsius
– Humidity - As a %
– WindSpeed - In m/s
– Season - Winter, Spring, Summer, Autumn
– Holiday - Holiday / No holiday
• Create a new variable called FullDate which includes the hour in it (set minute and second to zero).
• Change the factor levels of Holiday to Yes / No (use this order).
• Change the order of the Season factor levels to Spring, Summer, Autumn and Winter.
BikeWashingtonDC.csv
• Remove the following columns: unique record index, year, month, day of the week, working day, weather condition, normalised feeling temperature and number of bikes rented by casual and registered users (i.e. keep only the total count).
• Change the name of the columns to match the ones for Seoul.
• Convert Humidity to a %.
• Convert Temperature to degrees Celsius.
• Convert WindSpeed to m/s.
• Change the factor levels of Season to Spring, Summer, Autumn and Winter (in this order to match
Seoul’s one).
• Change the factor levels of Holiday to Yes / No (use this order).
• Create a new variable called FullDate which includes the hour in it (set minute and second to zero).
The Seoul and Washington data frame objects should now have the same set of consistently named and comparable columns.
Data visualisation
Next, explore (visually) the associations between bike usage, weather, time of day and holidays for both the Seoul and Washington datasets. Produce any number of relevant plots to answer the following questions, and comment on the similarities / differences between Seoul and Washington.
• How does air temperature varies over the course of a year?
• Do seasons affect the average number of rented bikes?
• Do holidays increase or decrease the demand for rented bikes?
• How does the time of day affect the demand for rented bikes?
• Is there an association between bike demand and the three meteorological variables (air temperature, wind speed and humidity)?
Statistical modelling
For both the Seoul and Washington datasets do the following:
• Fit a linear model with log count as outcome, and season, air temperature, humidity and wind speed as predictors. Print out a summary of the fitted models, comment on the results and compare across the two cities.
• Display the 97% confidence intervals for the estimated regression coefficients. Do you think these confidence intervals are reliable?
• Assuming the model is trustworthy, what’s the expected number of rented bikes in winter when the air temperature is freezing (0◦C), in the presence of light wind (0.5m/s) and a humidity of 20%. Provide the 90% prediction intervals and comment on the results. Hint: Use the interval argument of the predict function.