RPubs link information

https://rpubs.com/Raelene/681238

ACT Bike Barometer (Transport Canberra n.d.).

Introduction

• More physically active methods of transport, like bicycling, are becoming increasingly popular as a means to get to work, as well as for general fitness.
• Planning and building infrastructure to support these alternative methods relies on knowing the patterns of use and the uptake of these options.
• For example, how many cycle to work, are there an adequate number of bike racks or storage areas in destination sites, and are the bike paths sufficient for the numbers using them?

Introduction Cont.

• In the 2016 census, 1.1% of the total population reported bicycling to work, with 3.9% of people walking only, and Canberra saw the highest rate of people cycling or walking to work (8.4%; Australian Bureau of Statistics 2017).
• Canberra is popularly regarded as a bicycle friendly city and installed a bike barometer in November 2017 in O’Connor (a suburb close to the city center) to monitor and record bicycle use (Transport Canberra n.d.).
• Using the ACT bike barometer, can we establish a trend of more people bicycling to work, compared with bicycling for pleasure or exercise.

Problem Statement

• Specifically, are there more bicycle trips on working days (weekdays) compared with non-working days (weekends and public holidays).
• A box plot was used to compare weekday and weekend bicycle trips along with summary statistics.
• A two-samples t-test was conducted to test whether there was a significance difference in the average number of weekday and weekend bicycle trips.

Data

• The ACT bike barometer data set was sourced from the ACT Government’s Open Data Portal, accessed 28 September 2020 (ACT Government 2020).
• The bike barometer is located on the Sullivan’s Creek shared path intersection with MacArthur Avenue, O’Connor, ACT.
• The data recorded includes an hourly interval count of bicycle trips using a ground sensor, recorded as a DateTime variable (dd/mm/yyyy hh:mm)
  
• The number of bicycle trips is shown as the overall count (both directions), an inbound towards the city count, and an outbound away from the city count.

Data Pre-Processing

• The DateTime variable was separated into Date and Time variables in order to extract the year and month as new columns.
• The data was then filtered to only records for Autumn (Mar, Apr, May) and 2018, as the data set was very large and Autumn is the most temperate season in Canberra for outdoor activity.
• The date variable was then aggregated to give a daily count, rather than an hourly count, and renamed.

Data Pre-Processing cont.

• The weekday (i.e. Mon) was extracted from date.
• A new variable was then created as a factor which labeled days as either “Weekday” or “Weekend”, and was then reordered so “Weekday” was the first level.
• Public holidays were identified manually by checking dates and then recoded as “Weekend”. This data set was used for all analyses except checking the normality assumption.
• The normality assumption was check separately for weekday and weekend data by filtering the data beforehand.

Descriptive Statistics and Visualisation

• The average number of bicycle trips appears to be higher on Weekdays (m=1922.4, sd=258.7) compared with Weekends (m=683.2, sd=178.6). This is also reflected in the following boxplot.
• There were two suspected outliers detected in the Weekday data, these two observations were identified as being close to the school holidays and were retained as weekdays. No outliers are seen in the box plot for the Weekend data.

boxplot(Number_Trips ~ Wday_Wknd,
        data = bike_Autumn2018, main = "Box Plot of Number of Bicycle Trips by Day Type", ylab = "Number of Trips", xlab = "Day Type")

Descriptive Statistics and Visualisation cont.

bike_Autumn2018 %>% group_by(Wday_Wknd) %>% summarise(
  Min = min(Number_Trips, na.rm = TRUE),
   Q1 = quantile(Number_Trips, probs = .25, na.rm = TRUE),
  Median = median(Number_Trips, na.rm = TRUE),
  Q3 = quantile(Number_Trips, probs = .75, na.rm = TRUE),
  Max = max(Number_Trips, na.rm = TRUE),
  Mean = mean(Number_Trips, na.rm = TRUE),
  SD = sd(Number_Trips, na.rm = TRUE),
  n = n()) -> table1
knitr::kable(table1)

Hypothesis Testing - Normality

• The filtered Weekday data was tested for the assumption of normality using a QQPlot. While there were a couple of outliers (previously seen in the box plot), the rest of the data points are fairly closely following the line.
• The assumption that the underlying distribution for the data is normal is not violated.

bike_Weekday$Number_Trips %>% qqPlot(dist="norm", main = "Test for normality (Weekday)", ylab = "Number of Trips")

## [1] 43 48

Hypothesis Testing - Normality

• The filtered Weekend data was tested for the assumption of normality using a QQPlot. While there were two outliers, one at either end, the rest of the data points follow the normal distribution line quite closely.
• The assumption that the underlying distribution for the data is normal is not violated.

bike_Weekend$Number_Trips %>% qqPlot(dist="norm", main = "Test for normality (Weekend)", ylab = "Number of Trips")

## [1] 25 20

Hypothesis Testing - Homogeneity of Variance

• The assumption of equal variance was tested using the Levene’s test. The p-value was greater than the 0.5 significance level, so it is safe to assume equal variance.

leveneTest(Number_Trips ~ Wday_Wknd, data = bike_Autumn2018)

Hypothesis Testing - t-Test

• Assuming equal variance, the two sample t-test results showed a statistically significant difference between the number of bicycle trips on weekdays compared with weekends t(df = 90) = 23.90, p<.001, 95% CI for difference in means [1136.24, 1342.26].
• This suggests that number of bicycle trips on weekdays is significantly higher than on weekends.

t.test(
  Number_Trips ~ Wday_Wknd, 
  data = bike_Autumn2018,
  var.equal = TRUE,
  alternative = "two.sided"
)

## 
##  Two Sample t-test
## 
## data:  Number_Trips by Wday_Wknd
## t = 23.901, df = 90, p-value < 2.2e-16
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  1136.240 1342.257
## sample estimates:
## mean in group Weekday mean in group Weekend 
##             1922.4098              683.1613

Discussion

• Weekdays were found to have a significantly higher average number of bicycle trips than weekends, by up to almost three times.
• The results support the supposition that more people are bicycling as a means of transport for work than for other reasons.
• There were some outliers identified in the data, which may be improved by a re-examination of the data taking school holidays into account, possibly presenting them as a third group.
• The popularity of bicycling as a means of getting to work has been supported with the findings of this study and should be taken into account for infrastructure planning into the future.

References

ACT Government 2020, ACT Bike Barometer - MacArthur Avenue, data file, ACT Government, Transport Canberra and City Services, Canberra, viewed 28 September 2020, https://www.data.act.gov.au/Transport/ACT-Bike-Barometer-MacArthur-Avenue/62sb-92ea.

Australian Bureau of Statistics 2017, More than two in three drive to work, Census reveals, media release 133/2017, ABS, Canberra, viewed 22 October 2020, https://www.abs.gov.au/ausstats/abs@.nsf/mediareleasesbyreleasedate/7DD5DC715B608612CA2581BF001F8404?OpenDocument.

Transport Canberra n.d., Bike barometer and active travel data – Data collection and usage, ACT Government, Canberra, viewed 16 October 2020, https://www.transport.act.gov.au/about-us/active-travel/active-travel-in-the-community/cycling-in-canberra/Bike-barometer-and-active-travel-data.