1 Introduction
The evolving landscape of statistics and data science education increasingly emphasizes the critical role of computing. This shift [17, 28] reflects a broader trend towards integrating computational methods with traditional statistical techniques. These skills are indispensable in modern data analysis, enabling students to effectively handle and interpret complex datasets. A comprehensive teaching approach in data science [18] is further advocated, one that seamlessly blends computational tools with statistical concepts. The inclusion of programming languages like R [32], Python [38], and Julia [7] in their curricula is not just about tool proficiency; it is about fostering a deeper understanding of data analysis and statistical inference in a digitally driven world.
This growing emphasis on computing has led to practical educational events that give students the opportunity to apply their computational and statistical skills. Data competitions, such as hackathons [24] and DataFests [29], represent vital practical components in the education of future data scientists and statisticians. While hackathons typically focus on creating a working prototype or solution in a short time, often with a strong emphasis on coding and immediate problem-solving, DataFests usually involve more in-depth analysis of large datasets, focusing on statistical insights and data visualization. These events challenge participants to apply their skills in real-world scenarios, encouraging not just the application of statistical and computational knowledge but also the development of collaborative and innovative problem-solving abilities. However, a key limitation of these traditional events is that they typically require participants to already have a strong background in programming and data science, making it difficult for beginners or those with limited experience to fully engage.
Recognizing the limitations of traditional events, the Data Jamboree was developed as a more flexible educational experience, designed to accommodate participants with varying levels of expertise. While hackathons and DataFests also promote inclusivity, mentorship, and hands-on learning, Jamborees emphasize a workshop-style format that allows participants—especially those with less technical background—to engage fully from the outset. Led by experienced mentors, Jamborees foster collaboration across skill levels, ensuring that all participants can contribute meaningfully regardless of their prior experience in data science. The Data Jamboree format is flexible and could serve as a standalone event or as a component within larger events like DataFests, providing an accessible entry point for participants before engaging in more complex challenges. It incorporates open-source tools such as Julia, Python, and R, allowing participants to explore and compare the strengths of each in solving real-world problems. The event’s flexible structure can be adapted based on its goals and duration, promoting hands-on learning through active coding sessions, collaborative problem-solving, and guided practice in data science.
The Data Jamboree, organized by the American Statistical Association’s (ASA) Section on Statistical Computing and co-sponsored by the ASA’s Section on Statistical Graphics, has been held twice, first in November 2022 and again in 2023. As part of the annual mini-symposium “Statistical Computing in Action,” the event gathers data enthusiasts, professionals, and students to engage in data cleaning, manipulation, and analysis of open data projects. The Jamboree promotes the exchange of knowledge and skills among participants through hands-on experiences. The 2023 Jamboree was a one-hour event, the concise format of which was shaped by the symposium schedule, offering a focused yet impactful experience that emphasized hands-on learning and collaborative problem-solving. A unique feature of the event is its focus on using multiple open-source tools to collaboratively solve data science problems. For example, in 2023, participants used the New York City (NYC) Open Data’s 311 Service Requests (SR) dataset. The event provided opportunities to explore and evaluate different programming languages—Julia, R, and Python—enabling participants to enhance their practical skills while encouraging innovation, critical thinking, and teamwork [28, 13].
The use of open data in the Data Jamboree offers both educational and practical advantages. Open data initiatives promote transparency and public engagement, which are essential in data science [6, 20]. By working with open data, participants face real-world challenges, using Julia, Python, and R to tackle the same data science problems. This hands-on approach not only enhances technical skills but also highlights the societal impact of data science. Real-world datasets present the messy and complex issues that professionals encounter, offering more valuable learning experiences than curated or synthetic datasets often used in classrooms. The open data format allows participants to address meaningful issues such as public service improvement and urban planning. Additionally, it demonstrates the versatility of these tools, showing how each can be used to manipulate, analyze, and visualize data differently. The application of these languages in real-world scenarios supports the notion that open data can foster diverse, data-driven solutions [8, 10]. Furthermore, using open data illustrates how data science contributes to broader societal benefits [34]. Ultimately, the Data Jamborees serve as forums where participants use open-source software and open data to create valuable educational experiences.
The insights gained from the Data Jamboree provide a unique perspective on how diverse tools like Julia, Python, and R can be applied in educational and real-world settings. By evaluating the strengths and weaknesses of each tool, we offer guidance for educators and data science practitioners. The contributions of this article are three-fold. First, we report on the Data Jamboree held at the 2023 Mini-Symposium of the ASA Section on Statistical Computing, with detailed descriptions of the Jamboree data sets and task, including the cleaning, manipulation, and analysis of the NYC 311 SR data. Second, we discuss the use of three popular open-source data science environments—Julia, Python, and R—within the context of the Jamboree’s problems, evaluating their strengths and limitations. Third, we provide recommendations for designing engaging Data Jamboree events that promote data science outreach and education.
The rest of the paper is organized as follows. We first present the dataset and problems of the Data Jamboree at the 2023 Mini-Symposium on Statistical Computing in Action, which sets up the arena for the different data science languages in Section 2. A comparison of Julia, Python, and R with the major advantages and limitations of each is summarized in Section 3. Recommendations on how to design a Data Jamboree event are provided in Section 4. A discussion concludes in Section 5. The code for handling the Jamboree problems using all three languages is publicly available in GitHub repositories.
2 The 2023 Data Jamboree
The 2023 Data Jamboree attracted over 120 participants, with about half being students, including 49 student members from the ASA Section on Statistical Computing. The rest of the attendees were early-career professionals and experienced data scientists. Participants brought varying levels of familiarity with R, Python, and Julia, fostering a collaborative environment that encouraged both peer learning and skill development.
2.1 Data and Tasks
The NYC 311 SR dataset represents a comprehensive accumulation of non-emergency requests made by New York City residents. The dataset includes a wide array of urban issues, ranging from noise complaints and pothole reporting to graffiti removal and street light issues. Since its inception in 2010, the dataset has served as a valuable resource for understanding urban living dynamics and informing public service improvements. By providing insight into the spatial and temporal patterns of urban issues, the dataset not only reflects the day-to-day concerns of New Yorkers but also offers opportunities for analysis that are relevant to urban planners and policymakers [e.g., 27, 23, 2]. The open accessibility of this dataset makes it an ideal resource for data science projects and competitions, as it offers both real-world relevance and a wide scope for exploration.
For the 2023 Data Jamboree, a subset of the NYC 311 SR dataset was selected, consisting of requests created between January 15 and 21, 2023. The dataset was downloaded in CSV format from the NYC 311 Open Data Portal on February 22, 2023. This subset contained 54,469 rows and 41 columns, with a total size of 32 MB. Although the subset covers only one week, it still provided participants with real-time urban challenges, such as handling incomplete data, testing associations, and managing large datasets typical of urban environments. This dataset has also been utilized in various educational settings to address civic questions. For example, in STAT 3255/5255, Introduction to Data Science at the University of Connecticut, students explored the NYC 311 data for homework assignments and midterm projects. Selected students presented their analyses at a workshop during NYC Open Data Week in March 2023. While independent from the Data Jamboree, these educational applications highlight the dataset’s versatility and relevance for real-world data science problems.
The scientific tasks of the Jamboree were divided into three main stages:
Data Cleaning: In this stage, participants were required to standardize column names across programming languages for ease of comparison, identify and correct errors or inconsistencies (e.g., closed dates occurring before created dates, invalid values), and handle missing data (e.g., using geocoding to fill in missing zip codes). This stage also included a reflection task, where participants summarized their suggestions for improvements to the data, which were to be communicated to the data curator. These reflections ensured that participants not only engaged with data cleaning tasks but also considered broader data quality issues.
Data Manipulation: Focusing on New York Police Department (NYPD) requests, this stage provided specific instructions for participants to create a new variable, ‘duration’, representing the time span between the Created Date and the Closed Date. Participants were then requested to visualize the distribution of the ‘duration’ variable across weekdays, weekends, and boroughs, and to test for similarities in these distributions. Additionally, they were instructed to merge zip code-level data, such as population density, home values, and household income from the US Census, with the NYPD requests data for further analysis. This added dimension allowed for a more nuanced exploration of the socioeconomic factors influencing the handling of NYPD service requests. The structured approach ensured participants could successfully complete foundational tasks, setting the stage for more open-ended analysis in subsequent stages.
Data Analysis: The final stage required participants to define a binary variable, ‘over3h’, indicating whether the time to close a NYPD SR exceeded three hours. They were tasked with using logistic regression models to predict this variable, incorporating 311 SR data and zip code-level covariates. Participants were also encouraged to apply alternative models, such as random forests or neural networks, to explore different analytical approaches. This diversity of modeling techniques allowed participants to compare the performance and interpretability of various methods, enhancing their understanding of predictive modeling in real-world data contexts.
Additional details on the Jamboree’s tasks are available at https://asa-ssc.github.io/minisymp2023/jamboree/.
2.2 Workshops
The 2023 Data Jamboree featured three of the most widely used open-source programming languages in data science: Julia, Python, and R. Each language was selected for its strengths in different aspects of data science. Python, known for its versatility and broad array of libraries, is widely used in machine learning and general-purpose data manipulation [39]. R, on the other hand, is highly regarded for its statistical analysis and visualization capabilities [43], making it a favorite among statisticians and data scientists alike. Julia, with its growing user base, has gained popularity for its high performance in handling large datasets [7], positioning itself as a strong competitor in scientific computing and data-heavy tasks. These languages have been compared in various applications for their reproducibility and performance [e.g., 36].
The workshops were featured events during the Data Jamboree, and while attendance was voluntary, they were central to the event’s structure. Participants were advised to pre-install Julia, Python, and R ahead of the event to ensure smooth participation. However, installation of add-on packages was managed during the workshops, as all three languages provide convenient package management systems that allow for quick and easy installation. Each language was allocated 15 minutes to address the same tasks using the NYC 311 data. During these sessions, the leader presented their solutions on screen while participants followed along on their own computers using code shared via GitHub repositories. This follow-along structure ensured that participants could engage directly with the material during the workshop. The flexible nature of this format, as described in Section 4, allows organizers to adjust the event length and level of detail based on their time budget and participant engagement. After the demonstrations, a question-and-answer session allowed participants to explore the tools further.
The Julia, Python, and R workshops were led by HaiYing Wang, Shannon Taas, and Lucy McGowan, respectively. The Jamboree was moderated by Sam Tyner, who had led the R workshop during the 2022 Data Jamboree. Video recordings of the workshops are publicly available on the YouTube channel of the ASA Section on Statistical Computing and ASA Section on Statistical Graphics (https://www.youtube.com/@statgraphics). The codes used by the workshop leaders are also publicly available in GitHub repositories.
3 Comparison of Julia, Python, and R
The Data Jamboree provided an ideal platform to compare three popular open-source languages—Julia, Python, and R—within the context of real-world data science tasks. These tools were evaluated based on their performance in handling the NYC 311 SR dataset during the Jamboree, specifically focusing on data cleaning, manipulation, and analysis. The comparison takes into account each tool’s efficiency in managing large, complex data, their flexibility in communicating with other tools, and their power in performing statistical analysis and visualization. Furthermore, this evaluation considers the educational impact of these tools, particularly their accessibility for learners with different skill levels and their ability to support hands-on learning experiences in data science. By framing this comparison around the practical tasks completed by participants during the Jamboree, we highlight the strengths and limitations of Julia, Python, and R in both professional and educational settings.
3.1 Julia
The Julia solution to the Jamboree problems by HaiYing Wang is publicly available at https://github.com/Ossifragus/DataJamboree. Julia’s just-in-time (JIT) compilation offers advantages in processing large, complex datasets, making it highly efficient. For example, during the Data Jamboree, Julia efficiently managed the extensive NYC 311 SR data, performing aggregations and joins to analyze the distribution and frequency of various SR across different boroughs. The ability to compile code to machine language allows Julia to execute these tasks at speeds comparable to traditional high-performance languages like C and Fortran.
One of Julia’s key strengths is its flexibility in integrating with other tools, allowing participants to access the powerful capabilities of Python and R. Through the PythonCall package [35], Julia can call Python’s machine learning libraries, such as TensorFlow and Scikit-learn, which was especially useful during the Data Jamboree when we analyzed the NYC 311 SR data. This enabled them to apply advanced machine learning models for predicting trends in SR, while leveraging Julia’s computational efficiency for handling large datasets. Additionally, Julia’s RCall package RCall.jl provides access to R’s specialized statistical methods and visualization tools, such as ggplot2 [42], further enhancing the analysis of temporal and spatial patterns in the data. By integrating Python’s machine learning capabilities and R’s statistical tools, Julia offered participants a versatile environment to perform complex data manipulation, analysis, and visualization tasks on the NYC 311 SR dataset.
Julia also provides a first-class reproducibility system that stands out compared to R and Python, thanks to its native Pkg system [22]. Unlike external tools in other languages, Julia’s package management and environment isolation are deeply embedded within the language itself, with a strong focus on reproducibility from the ground up. The Pkg system ensures every project operates in its own isolated environment, with specific package versions that are fully insulated from other projects. This greatly minimizes dependency conflicts and makes it easy to recreate identical environments across different systems, enhancing both reliability and collaboration.
In terms of data manipulation and computational tasks, Julia offers powerful capabilities through packages like DataFrames.jl [9]. These libraries provide comprehensive functionalities for handling large datasets and performing complex transformations essential for exploratory data analysis. During the Data Jamboree, we utilized DataFrames.jl to efficiently filter, sort, and aggregate the NYC 311 SR data. For instance, we analyzed the temporal trends of different types of SR and identified peak times for noise complaints, enabling more precise analysis and reporting.
However, Julia’s ecosystem is not as mature as that of Python or R. The availability of specialized libraries and community support for niche areas can be limited. During the Data Jamboree, we faced challenges finding well-documented libraries for specific tasks, such as geospatial analysis, which are readily available in Python or R. This limitation can slow down the workflow, requiring users to develop custom solutions or bridge Julia with other languages.
Additionally, Julia has a steeper learning curve compared to Python. While praised for its performance, new users had to spend additional time learning its syntax and understanding its ecosystem. This initial learning investment can be a barrier in time-constrained environments like hackathons, where quick onboarding and rapid development are crucial. Novices could struggle with the syntax differences when transitioning from more familiar environments, impacting their productivity initially.
Integrating Julia into existing data science workflows that predominantly use Python or R can also present challenges. Despite tools like PyCall [21] and RCall [4] facilitating integration, differences in syntax, data handling, and library functionalities require additional adjustments. During the Data Jamboree, we needed to create wrappers and compatibility layers to ensure smooth interoperability between Julia and other languages. For instance, integrating visualization outputs from Julia with a dashboard created in Python’s Dash [30] required careful handling of data formats and function calls, introducing complexity and potential points of failure in the workflow.
In conclusion, Julia’s performance during the Data Jamboree highlighted its potential and limitations. Its ability to efficiently handle large datasets, flexibility in integrating with other tools, and strong capabilities for statistical analysis and visualization make it a powerful tool for data science. However, the relatively immature ecosystem, steeper learning curve, and integration challenges with established workflows are notable drawbacks. As Julia continues to evolve, it is poised to become a more integral part of the data science toolkit, particularly for tasks demanding high computational performance.
3.2 Python
The Python solution to the Jamboree problems by Shannon Tass is publicly available at https://github.com/esnt/Jamboree. Python’s extensive library ecosystem and user-friendly syntax made it a popular choice among participants for various data science tasks. The Jamboree showcased Python’s strengths in data manipulation, integration with other tools, and data visualization, while also revealing some of its limitations.
Python’s versatility is one of its key strengths, supported by a rich ecosystem of libraries designed for data manipulation, analysis, and visualization. During the Data Jamboree, we leveraged Pandas [26], a powerful data manipulation library, to handle the NYC 311 SR data efficiently. Tasks such as data cleaning, filtering, and aggregation were streamlined using Pandas’ intuitive syntax. For example, participants used Pandas to filter SR by borough and service type, enabling a detailed analysis of complaint patterns across different areas of New York City.
In addition to data manipulation, Python excels in integrating with various tools and environments, enhancing its utility in diverse workflows. The ability to call R functions using the rpy2 package [16], for instance, allowed participants to utilize R’s statistical packages within a Python environment. This was particularly useful for performing advanced statistical tests on the NYC 311 SR data, where Python was used to handle the data preparation and visualization, and R was used to conduct the statistical analysis. Such integration capabilities underscore Python’s flexibility in combining strengths from different programming languages.
Python’s data visualization capabilities are another advantage, with packages like Matplotlib [19] and Seaborn [41] providing tools for creating detailed and informative visualizations. During the Data Jamboree, we used these libraries to visualize trends and distributions in the NYC 311 SR data. For instance, Seaborn was used to create heatmaps showing the density of SR in different neighborhoods, while Matplotlib helped in plotting time series to analyze trends in complaint types over time. These visualizations were crucial for identifying patterns and drawing actionable insights from the data.
However, Python also has its limitations. One of the challenges one faces is performance issues when handling extremely large datasets. While libraries like Pandas for data frames are powerful, they are not optimized for very large-scale data processing, leading to slower performance compared to languages like Julia. For example, aggregating millions of records in the NYC 311 SR dataset sometimes resulted in significant processing times, highlighting the need for performance optimization techniques or the use of more specialized tools like the package Dask [14].
Another limitation is Python’s less mature support for certain statistical and mathematical operations compared to R. While Python’s libraries are comprehensive, they may not offer the same depth of specialized statistical functions available in R. During the Jamboree, we occasionally needed to switch to R for specific statistical analyses that were more straightforward to perform using R’s specialized packages. This dual-language workflow, although effective, added complexity to the process.
Python’s learning curve, while generally considered accessible, can still pose challenges for complete beginners. Participants new to programming or data science had to invest time in learning Python’s syntax and understanding its diverse library ecosystem. Despite its readability and user-friendly nature, the initial learning phase can be daunting, especially in a high-pressure environment like a hackathon.
In conclusion, Python’s performance during the Data Jamboree highlighted its versatility and extensive library support for data manipulation, integration, and visualization. Its ability to combine strengths from different programming languages through seamless integration makes it a powerful tool for data science. However, performance limitations with very large datasets, less mature statistical support compared to R, and the initial learning curve are notable drawbacks. As Python continues to evolve, it remains a fundamental tool in the data science toolkit, particularly for tasks requiring extensive data manipulation and visualization.
3.3 R
The R solution to the Jamboree problems by Lucy D’Agostino McGowan is publicly available at https://github.com/LucyMcGowan/2023-data-Jamboree. The Jamboree highlighted R’s powerful data manipulation tools, integration with other languages, and advanced visualization capabilities, while also revealing some of its limitations.
R’s strength in data manipulation was evident as we used the dplyr [46] and tidyr [44] packages to clean and transform the NYC 311 SR data. These packages allowed for streamlined data wrangling, making it easy to filter, sort, and aggregate the data. For example, we used dplyr to group SR by type and borough, summarizing the frequency of different complaint types across New York City. The tidyverse suite [45], which includes dplyr and tidyr, provides a cohesive set of tools with a shared syntax and grammar that simplify complex data manipulation tasks, making R a powerful language for data scientists.
Another advantage of R is its advanced statistical analysis capabilities. R has a rich ecosystem of packages for performing a wide range of statistical tests and models. During the Data Jamboree, we used standard packages like stats and MASS [40] to conduct hypothesis testing and fit regression models. For instance, we analyzed the correlation between complaint types and response times, using R’s statistical tools to derive meaningful insights from the data. R’s extensive collection of widely used specialized packages, such as survival [37] for survival analysis and lme4 [5] for mixed-effects models, enables users to perform advanced statistical analyses that are crucial in many data science projects.
R’s visualization capabilities are another area where it excels. The ggplot2 package [42], also part of the tidyverse, is renowned for its ability to create high-quality, customizable visualizations. We used ggplot2 to create detailed plots that showcased trends and patterns in the NYC 311 SR data. For example, we generated heatmaps to visualize the density of SR in different neighborhoods and time series plots to analyze the trends in complaints over time. These visualizations helped us identify patterns and communicate the findings effectively.
However, R also has its limitations. One challenge is performance when handling very large datasets. While R is highly efficient for many data manipulation tasks, it can struggle with memory management and processing speed for datasets approaching one-tenth of computer memory. For the 2023 Data Jamboree tasks, one could experience slow performance if working with the full NYC 311 SR dataset, necessitating the use of data sampling or external data storage solutions to manage memory usage effectively.
Another limitation is R’s integration with other programming environments. While R can call functions from other languages using interfaces like reticulate [12] for Python and Rcpp [15] for C++, these integrations can sometimes be cumbersome. One could face challenges when integrating R with external tools or other programming languages, adding complexity to their workflows.
Lastly, R can have a steep learning curve for users who are new to programming or data science. While R’s syntax is highly expressive and well-suited for statistical analysis, it can be less intuitive for beginners. The tidyverse suite of packages seeks to bridge this gap, but many users have to supplement their analyses with other packages with differing syntax. Those who are new to R may have to invest additional time learning the syntax and understanding its extensive package ecosystem, which could be a barrier in fast-paced hackathon environments.
In conclusion, R’s performance during the Data Jamboree highlighted its strengths in data manipulation, statistical analysis, and visualization. Its cohesive set of tools and specialized packages make it a powerful language for data scientists. However, challenges with performance on large datasets, integration with other environments, and a steeper learning curve are potential drawbacks. As R continues to evolve, it remains an essential tool in the data science toolkit, particularly for projects requiring advanced statistical analysis and high-quality visualizations.
4 Design Your Own Jamboree
Designing a successful Data Jamboree requires careful planning and several key considerations to ensure an engaging and productive experience for participants. Below are essential guidelines to help structure the event effectively.
Defining the Aims The first step is to clearly define the objectives of the event. Identify your target participants—students, data professionals, or a combination of both—and determine what they should gain from the event. Will the focus be on general education, skill-building, or performance comparisons across different tools? Establish whether the Jamboree will emphasize statistical analysis, predictive modeling, or data visualization. These decisions will guide the event’s structure and content. The Data Jamboree format is highly adaptable and can function as a standalone event or as a featured component within larger data science gatherings, such as DataFests. Integrating a Jamboree into a DataFest could serve as a preparatory workshop, providing participants with hands-on experience before engaging in more complex data challenges.
Selecting the Data Choosing the right datasets is pivotal to the success of the Jamboree. Leveraging open data initiatives provides access to diverse and meaningful datasets. Examples include NYC Open Data, which offers a wide range of public datasets from various city agencies, data.gov for nationwide datasets across different sectors, NOAA for comprehensive climate data, and the U.S. Census Bureau for demographic data. A well-chosen dataset that is rich and complex can support a variety of data science tasks, such as data cleaning, curation, and statistical analysis. Utilizing datasets from multiple sources introduces additional challenges, particularly in terms of data integration and manipulation, which mirrors real-world scenarios. This complexity not only enhances participants’ technical skills but also provides a more comprehensive understanding of data science workflows.
Designing the Jamboree Questions The questions posed to participants should cover a range of data science tasks, such as data cleaning, exploratory analysis, statistical testing, predictive modeling, and visualization. This ensures that participants can apply a wide range of skills during the event. For example, tasks may involve identifying and rectifying data errors, exploring trends in the data, conducting statistical tests, or building machine learning models to predict outcomes. These questions should be aligned with the event’s overall aims and the complexity of the chosen data.
Determining the Event Length The event length should align with the complexity of the tasks and the organizers’ time budget. While our Jamboree was completed in one hour, higher levels of participant engagement, dedicated software setup time, extended question and answer sessions, or more in-depth tasks—such as machine learning or predictive modeling—may necessitate a half-day or even a full-day event. The leader presentations, and participant activities are integrated into a single session, with participants following along as the leaders present solutions. This format provides flexibility, allowing organizers to adjust both the duration and depth of the event based on available time and audience needs. For longer events, logistical considerations like meal breaks and lodging become important, especially for participants traveling from afar. Ensuring access to essential facilities, including a reliable internet connection and adequate power outlets, is critical for maintaining productivity and focus. In virtual events, certain logistical challenges are reduced, but collaboration and communication can be more difficult, potentially affecting engagement. The duration should still reflect the difficulty of the tasks to ensure a balance between engagement and content coverage.
Finding Workshop Leaders Experienced workshop leaders play a pivotal role in ensuring the success of the Jamboree. Leaders should have expertise in both data science and the specific tools being used, such as R, Python, or Julia. They should guide participants through the tasks, offering insights and fostering collaboration. We recommend having at least one leader per programming language or key task, with a student-to-instructor ratio that ensures adequate support for all participants, including teaching assistant if resources allow. Leaders may come from academic institutions, industry, or data science communities.
Special Considerations for Minors If minors are involved, additional steps must be taken to ensure their safety and compliance with legal requirements. This includes obtaining parental consent, designing age-appropriate challenges, and providing adequate supervision. Any legal or institutional requirements must be adhered to, such as ensuring leaders have undergone background checks or assigning designated guardians during the event. Addressing these considerations helps create a safe and welcoming environment for younger participants.
Post-Event Analysis and Feedback Post-event analysis is essential for refining future Jamborees. We recommend distributing surveys to gather feedback from participants about their experience. To maximize response rates, feedback forms should be administered immediately after the event. Additionally, holding a debrief session for leaders allows for the identification of what worked well and where improvements can be made. Documenting these insights will help establish a foundation of knowledge that can be shared with future organizers. Over time, this feedback will contribute to an evolving institutional knowledge that enhances the quality of the event and ensures its continued success.
These recommendations provide a comprehensive framework for organizing a successful Data Jamboree. By focusing on clear objectives, appropriate datasets, well-designed questions, event logistics, and thoughtful post-event analysis, organizers can create an engaging and educational experience for participants. Ensuring a balance between structure and flexibility fosters both innovation and collaboration.
5 Discussion
Comparing Julia, Python, and R, we find each language has unique strengths and limitations that impact their effectiveness in a data science educational setting. Julia stands out for its high performance, particularly with large datasets, due to its just-in-time (JIT) compilation. Additionally, Julia can seamlessly call Python libraries using the PyCall package [21] and R libraries using RCall [4], enhancing its versatility. Python is celebrated for its extensive library support and accessibility, making it powerful for a wide range of applications, particularly in deep learning with libraries like TensorFlow [1], Keras [11], and PyTorch [31]. R excels in statistical analysis and visualization, with a comprehensive ecosystem of packages tailored for these purposes. Unified data formats, such as Arrow [3], can further streamline cross-platform data manipulation and enhance computational efficiency across Julia, Python, and R.
Notably, we did not include commercial software such as SAS, Matlab, or Mathematica in our comparison. These tools, while powerful, are less accessible due to licensing costs and do not align with the open-source philosophy that fosters broad educational engagement. Furthermore, we do not recommend tools like MS Excel for data science. While Excel can serve as an accessible entry point for basic data manipulation and visualization, it may present challenges for advanced data analysis due to its limitations in accuracy and susceptibility to user errors [25, 33]. Additionally, reliance on Excel may encourage poor practices such as manual data manipulation, lack of reproducibility, and overuse of hard-coded values, which hinder the scalability and transparency of analyses. Time invested in learning Excel could be more effectively used to master specialized tools like Julia, Python, and R, which offer greater capabilities for data science.
The Data Jamboree format is broadly applicable to events like DataFests and other educational activities, emphasizing the value of tackling real-world problems with open data. Unlike hackathons, the workshop style of the Data Jamboree encourages participation from all skill levels, not just those with coding experience. Using real-world datasets, such as those provided by Open Data initiatives, allows participants to tackle practical challenges, enhancing their problem-solving skills and understanding of data science applications. This approach not only fosters engagement and excitement but also prepares students for professional scenarios where they must analyze, interpret, and make decisions based on real-world data. Incorporating these elements into educational events ensures a comprehensive learning experience that bridges the gap between theoretical knowledge and practical application.
To further enhance the educational impact of future Data Jamborees, incorporating systematic performance evaluation and participant feedback will be valuable. While the 2023 Data Jamboree did not focus on detailed profiling of computational performance across programming languages, future events could incorporate systematic benchmarking as part of the analysis. This would provide participants with deeper insights into the efficiency and scalability of different tools, particularly when working with larger datasets or more complex tasks. Future Data Jamborees could also benefit from systematically collecting data on participant task completion rates and experience levels. This information would offer valuable insights into how different audiences engage with the material and could guide adjustments in pacing, content depth, and workshop structure to better meet participant needs.
