data set for statistics project

data set for statistics project

Optimizing Data Set Selection for Statistics Projects

1. Introduction to the Importance of Data Set Selection

Researchers in statistics education look at problems related to how to teach statistical concepts effectively. Teaching with data promotes deep learning by demonstrating the relevance of statistical concepts and methods and motivating questions. In this research project, rather than developing new teaching methods, promoting better teaching, or supporting student learning, we examine differences in student outcomes in statistically heavy courses. Unlike previous research, we are not comparing two particular teaching methods or looking at the potential of innovative projects. Instead, we examine a key aspect of over a hundred statistically driven satisfaction with the data and how this impacts the course and student grades. From examining these differences in outcomes, our intention is to determine what dataset characteristics may inform the dataset choices of instructors. With this information, we are positioned to help instructors select the best available options for the constant flow of statistics projects.

Since the beginning of the field of statistics over a hundred years ago, statisticians have been developing new ways of analyzing large numbers of observations to understand real-world problems. Students who learn statistical techniques over the course of their education complete a variety of coursework-specific projects which require data obtained from real-world settings to analyze. It is interesting to consider this learning experience through the lens of big data. There are countless datasets to consider for use in such projects. Projecting research requires decisions on what dataset is best for student learning. There are many variables related to their learning experience to consider. What characteristics of a dataset make it optimal for the purpose of the learning outcomes of the course? By better understanding impacts on student achievement in statistics, we can help students to succeed.

2. Factors to Consider When Choosing a Data Set

First, consider what data collection is emphasized in the statistical process. Fundamental to the practice of statistics is the systematic data collection process, whether it be through a defined experiment, a survey, an observational study, or some form of modeling. The data set used in the project should provide students with a clear understanding of this process. While this may eliminate the collection of experimental data for some research questions, it provides an opportunity for students to understand principles of observational studies and for students to deal with confounding variables. It also highlights the role proper data collection plays in the accuracy and precision of estimates and the power of hypothesis tests. Students who have invested considerable time and resources in data gathering activities have a greater appreciation for the exploratory and analysis techniques employed later. Additionally, and possibly more crucial, is that the collected data often have more relevance and importance than that of a simulated data set or one obtained from a repository. For the research topics for which students were involved in data collection, they had a vested interest in the outcome of the analysis and subsequently showed more accountability during the production of the final report.

Once the final project is established, students are eager to begin the statistical analysis. For some students, their immediate impulse is to collect or use from the internet the largest data set available. Certainly, using large data sets offers some advantages, yet it also brings challenges and limitations. In this paper, we offer guidelines for the selection of data sets that will allow students to better engage in the statistical process while promoting sound analysis and results. We provide these guidelines in the context of a semester-long project in an undergraduate non-calculus-based introduction to statistics course, yet many of these guidelines can be applied to a term project in a laboratory component of a traditional calculus-based course.

3. Commonly Used Data Sets in Statistics Projects

– Datasets.gov: This site provides data set collections compiled by government agencies. – UCI machine learning repository: A collection of databases, domain theories, and data generators used to test machine learning algorithms. – IMF eLibrary: This financial data source covers a range of global financial data topics, including trade, debt, employment, and more. – CDC Data and Statistics: This is a collection of data sets provided by the Centers for Disease Control and Prevention. – National Center for Health Statistics: This is another CDC resource that provides downloadable public use data files containing information collected by NCHS. – WUKSACHI: The WUKSACHI data set repository contains a number of interesting statistical data sets to explore with tools like SAS JMP. – Rdatasets: This is a collection from the R Stats package often used by statisticians. – Statisti.ca: This data set includes a wide assortment of statistical databases covering income, income inequality, and more. – IMF Country Statistical Data: This is another economic statistics source from the International Monetary Fund. – fivethirtyeight: This collection provides curated data sets for R users of the R statistics package, including baseball data, homicide rates, Democrats and Republicans statistics, and much more.

The following data sets are commonly used in statistics projects:

4. Best Practices for Cleaning and Preparing Data Sets

A “code sheet,” which is a one-page description of everything you know about each variable in your dataset, is a document that you should keep religiously updated because almost anything you can write on it will save you a bit of time in the future. If you intend to use statistical packages like SPSS or SAS, there are certain best practices to which you should adhere. Of your dataset took responsibility for the age variable, someone else worked on the location assignment, and still someone else checked the entries on bald, both balding and not bald. In going through these exercises. And issues of survey or project design that are likely to pass some of the responsibilities we discuss during data collection over to the researcher analyzing the data, so in this chapter, we are going to emphasize the role that good preparation of a dataset plays in good data analysis.

While a detailed discussion of the cleaning and preparation of social research datasets is beyond the scope of this chapter, it is essential to point out that there are good practices in this area. If the ranges of revealed values in a dataset get into the habit of using, your project restrict. But at least initially, present a picture of what the ranges of data look to be because that might help reveal errors. Many researchers using survey data will go so far as to estimate the range of variation of each of their variables and write that information down on their “code sheets.”

5. Case Studies and Examples of Effective Data Set Selection

1. Conjoint analysis Companies, organizations, or industries can use survey methods to determine what customers want in a product or service but recognize such survey results can improve by presenting products that include those preferred attributes in combination. Develop the product(s) of interest and apply a fractional factorial design to determine which product combinations offer statistically significant appeal compared to a representative product with average attribute levels. For example, meat producers can ask if customers are willing to pay more for a hamburger or a steak to be described as ‘organic’, ‘grass-fed’, or ‘humanely raised’. A follow-up analysis could explore individual differences among segments.

This chapter contains 10 case studies demonstrating the role of data set selection in ensuring that evidence for claims in statistics projects is relevant. The case studies are drawn from quantitative methods, public policy evaluation, and data science. Data set selection is not limited to a single phase of the project timeline. Its design implications are evident in several phases including research question formulation and hypothesis testing. The author encourages that data set selection be taught throughout the data science curriculum and not just in a single class or project component. This chapter also describes two variations of effective data set use in undergraduate statistics courses in the form of a multistep modeling project to improve students’ observations of real-world phenomena and a data set selection teaching case study.