Introduction to Computer Science

Assignment Proposal
Department of Statistics 2025

Group Name:
1. Yosef Zinabu ----- UGR/3666/17
2. Yilkal Akele ------- UGR/1707/17
3.
4.
5.

Submitted to: Dr. Getahun

Date of submission: Nov 20,2025

 Proposal: Statistical Analysis of Ride (Ride App)
Addis Ababa

This proposal outlines a step-by-step plan to collect, clean, analyze, and

report statistical findings using trip-level and aggregated data from Ride (the
Ride-hailing platform operating in Addis Ababa). The project will address the
lecturer's 10 assignment topics (descriptive statistics, correlation,
comparative analysis, regression, time series, hypothesis testing, outlier
analysis, categorical analysis, ANOVA, chi-square). Deliverables include a
written report, Jupyter notebooks, charts, and a presentation.

1. Objectives
1.1 General objective

To perform a complete statistical analysis of Ride’s trip data to satisfy the

assignment requirements and produce actionable insights for the company
and stakeholders.

1.2 Specific objectives

 Obtain an anonymized dataset or suitable public/synthetic proxy

representing Ride trips in Addis Ababa.

 Produce descriptive statistics (mean, median, mode, range, variance,

SD) for key variables.

 Measure correlations and relationships between key variables

(distance, fare, duration, rating).

 Test differences across groups (time-of-day, vehicle types, promotion

vs non-promotion) using t-tests and ANOVA.

 Build predictive models to estimate fare and short-term demand

(regression and time-series models).

 Detect and treat outliers and assess their impact.

 Conduct categorical analyses and chi-square tests (payment method

vs satisfaction, vehicle type vs cancellation).

 Document methods, results, and code for reproducibility.

 2. Research questions
1. What are the mean, median, and mode for Ride’s fare, distance,
duration, and rating?

2. What are the range, variance, and SD of numerical variables in Ride’s

dataset?

3. Is there a significant correlation between fare and distance or duration?

Between surge multiplier and fare?

4. Are there differences in mean fares/durations between vehicle types,

time blocks, or promotion vs non-promotion trips?

5. Can we predict fare or demand using Ride’s available variables and

how much variance is explained?

6. What are the trends and seasonal patterns in daily trips and revenue
for Ride?

7. Did any recent Ride intervention (promotions, surge policy change)

significantly change trip counts or average fare?

8. Are there outliers in Ride’s data and how to handle them?

9. What are the proportions of vehicle types, payment methods, and

cancellation rates?

10. Is there an association between vehicle type and driver rating, or

payment method and satisfaction?

3. Data requirements and variables (About Ride)

3.1 Preferred trip-level scheme (Ride export or API)

 Trip ID - unique identifier

 Date & time - trip start timestamp (YYYY-MM-DD HH:MM)

 Provider - "Ride" (constant)

 Pickup zone - district/neighborhood (anonymized)

 Dropoff zone - district/neighborhood (anonymized)

 Distance km - trip distance

  Duration (min) - trip time in minutes

 Fare ETB - fare charged

 Base fare - base charge

 Surge multiplier - numeric factor

 Driver rating - rating by passenger (1–5)

 Vehicle type - sedan, minibus, moto, etc.

 Payment method - cash, mobile money, card

 Tip ETB - tip amount (if available)

 Cancellation flag - 0/1

 Promotion flag - 0/1 (discount applied)

 City zone code - coarse geocode for spatial aggregation (if allowed)

3.2 Aggregated fields (if trip-level not available)

 Daily trips, daily revenue, avg fare day, avg distance day, peak hour
trips

4. Data sources & collection methods

4.1 Primary: Request from Ride

 Formal data request: Prepare and send a polite, signed request to

Ride’s operations/data team asking for anonymized trip data for
educational use. Provide details on variables needed, anonymization
requirements, intended analyses, and deliverables.

 Data sharing options: CSV export, temporary API access, or

aggregated reports.

Template request items to include: purpose, date range desired (e.g.,

last 6 months), anonymization guarantee, contact details, and how results
will be shared.

4.2 Secondary options

 Public/similar datasets: Use open ride-hailing datasets (Kaggle,

GitHub) and adapt analyses to Ride’s context. We will annotate where
synthetic or proxy data were used.
  Manual/field collection: Short driver/customer survey for categorical
variables not in logs (consent required).

5. Sampling plan & sample size

 Preferred: 200+ trip records covering at least 1 months (to capture
weekly seasonality).

 Minimum: 150–200 records to permit meaningful inference.

 Sampling scheme: Random sampling across dates and hours;

stratified by vehicle type and city zone to preserve representativeness.

6. Data cleaning & preprocessing (Ride-specific steps)

1. Date/time parsing: Convert date time to datetime and extract hour,
weekday, and whether it is a weekend.

2. Fare per km: Compute fare per km = fare ETB/ distance km (handle
zero-distance with NaN).

3. Standardized categories: Normalize vehicle type labels, payment

method names.

4. Aggregate series: Build daily/weekly aggregates for time series

modeling.

7. Analysis plan (step-by-step, Ride-specific)

All tests use significance level α = 0.05 unless otherwise noted.

7.1 Descriptive Statistics

 Compute mean, median, mode, range, variance, SD, IQR for fare ETB,
distance km, duration min, driver rating, and tip ETB.

 Visuals: histograms and boxplots for fare and distance; bar chart for
vehicle-type counts.

Output: Table of descriptive stats with short interpretation (e.g., skewness

indicates right-skewed fares)

7.2 Correlation & Relationships

  Pearson correlation between fare ETB and distance km; between fare
ETB and duration min; Spearman if non-normal.

 Correlation between surge multiplier and fare ETB to quantify surge

effects.

 Visuals: scatter plots with fit lines; a correlation heatmap.

7.3 Comparative Analysis

 Compare average fare during daytime vs nighttime using t-test.

 Compare mean fares across vehicle type using ANOVA.

 Compare promotion flag vs non-promotion mean fares and trip counts.

7.4 Regression Analysis

 Simple model: fare ETB ~ distance km

 Multiple linear regression: fare ETB ~ distance km + duration min

+ surge multiplier + vehicle type + is weekend + hour of day

 Diagnostics: Check VIF for multicollinearity, residual plots for

heteroscedasticity, and normality of residuals.

 Consider log-transform of fare if skewed; use robust regression if

outliers impact results.

7.5 Time Series Analysis

 Aggregate to daily trips and daily revenue.

 Plot trends and decompose series (STL) into trend, seasonality,

residuals.

7.6 Hypothesis Testing (Intervention Analysis)

 If Ride implemented a promotion or policy change during the study

period, define pre/post windows.

 Use interrupted time series analysis or two-sample t-test (daily

aggregates) to test for level or trend change.

 Compute Cohen’s d for effect size.

7.7 Outlier Analysis

  Use distance to identify influential observations in regression.

 Compare model estimates with and without outliers; document effects.

7.8 Categorical Data Analysis

 Proportions of vehicle type, payment method, cancellation flag.

 Cross-tabulations: payment method vs driver rating categories; vehicle

type vs cancellation flag.

7.9 ANOVA (multi-group comparisons)

 One-way ANOVA: compare avg fare across multiple vehicle type

groups.

 Practical implication: which vehicle type is most cost-effective/popular.

7.10 Chi-Square Test

 Contingency table: vehicle type × rating category (e.g., 1–2: low, 3:

neutral, 4–5: high)

 Run chi-square of independence; check expected counts; use Fisher’s

exact if needed.

 Make simple predictive statements (e.g., based on past proportions,

expected percent of high-rated trips for a vehicle type).

8. Potential limitations & mitigations

 Data access denial: Use public ride-hailing datasets and clearly mark
limitations.

 Small sample/time window: Use non-parametric tests and clearly

discuss reduced statistical power.

9. Example hypothesis test (promotion impact)

 Example: Testing the impact of a 2-week promotion

o mean daily trips before promotion = mean daily trips after

promotion

o mean daily trips after promotion > mean daily trips before
promotion
o Test: two-sample t-test on daily trip counts (or interrupted time
series if data allows)