The Statistics Department instructional program has four primary targets:
- thousands of undergraduates taking a first, introductory course in statistics;
- hundreds of undergraduates taking intermediate level courses that involve data analysis;
- undergraduate and graduate students learning more advanced methods;
- graduate students developing data science skills for research.
In addition, there is a fifth, new audience that plan to target:
- non-traditional students seeking post baccalaureate training in statistical science.
The instructional plans connect targets to the themes of modern statistical inference. Students at all levels need basic lab skills of data handling, analysis and visualization. Intermediate and advanced students need to learn how to use, and possibly derive, models and methods for big data sets with complicated structures. Research in many fields will benefit from developing advanced skills for efficiently implementing new or hybrid methods. We are in the process of revising our curriculum to meet these needs. The Statistics Department had a productive retreat in January 2012 focused on our curriculum. We will follow this up with a more detailed look at course sequences later this year. Below are some goals to meet our instructional mission.
- Introductory Courses
- Statistical Languages
- Undergraduate Program
- Graduate Programs
- Non-traditional Subjects
- Non-traditional Programs
- Flexible Instruction
Introductory Courses: Develop a more uniform and engaging presentation of introductory courses, making them appealing and compelling and much less forbidding. Provide leadership for Lecturers of courses, now aimed at biological sciences (371), engineering (224 and 324) and a more diverse, general audience (301). Develop pedagogical tools to enable consistent instruction and to train both Lecturers and TAs. Renumber introductory courses to reflect level and provide obvious placement in course numbering. Explore how online or “blended” instruction can reach a wider audience of introductory students more efficiently. This process will be overseen by a new academic staff position that bridges between instruction and advising.
Statistical Languages: Provide campus leadership in training of statistical language concepts and tools, including best practices for introductory, intermediate and advanced audiences. The primary course delivery mechanism is a new modular, 1-credit course on “Learning a Statistical Language” (Stat 327, temporarily 692), which complements CS 368. Stat 327/692 began in Spring 2012 with “Introductory Data Analysis with R”, but will expand to other statistical languages (Stata, SAS) as demand and instructional resources emerge. Intermediate and advanced modules of Stat 327 will focus on reproducible research practices, such as annotating graphics, building functions and packages, and documenting work, useful in many advanced undergraduate and graduate statistics courses. We expect to offer modules every semester, aimed at undergraduate statistics majors, students taking a second course in statistics below 600, and new graduate students in statistics taking our core linear models series, Stat 849-850. This course will be overseen by a new academic staff position that bridges between instruction and computing.
Undergraduate Program: Increasing numbers of undergraduates across the nation are majoring in statistics. Why? As John Tukey said, “The best thing about being a statistician is that you get to play in everybody's backyard.” This surprising shift forces us to streamline and rationalize upper undergraduate course offerings to provide coherent, partially ordered sequences leading to a certificate or major in statistics. We need to include new opportunities for research experience with big data, emphasizing both methodological concepts and use of statistical language tools. An excellent beginning was accomplished this year. We will explore how modular offerings can increase flexibility for students and instructors to build a world-class undergraduate program.
Graduate Programs: Companies, government agencies and academic institutions want more MS and PhD graduates than our field can presently produce. Our MS offers world-class training that prepares our graduates to immediately contribute in a new workplace, because we emphasize both technical and communication skills in our training. Our PhD program has yielded leaders in all levels of academia, industry and government. This past success was built on a long history of innovation in instruction. We are due for another major shift to bring modern statistical research methods more firmly into the curriculum. Both MS and PhD programs need curricular overhaul to condense key training and free up time for big data research experiences. In particular, PhD students should have less classroom time and more opportunities with research data, leading to early publications. Instruction of linear models needs to be redesigned to efficiently meet training needs for MS (more applied) and PhD (more theoretical) students. We need to revisit our PhD qualifier to reassess its goal and focus material on advances in the field that we expect for our graduates. Simplifying graduate elective courses to broad topic areas, such as large sample inference, model selection, and dependent data, will enable creation of multiple modules per semester that focus on subjects that reflect student and faculty interest, to enhance entry into dissertation research.
Non-traditional Subjects: Our department, in collaboration with other units on campus, has research expertise in several subjects that are “marketable”. We could enhance our traditional programs by offering new (or in some cases retooled) courses in these areas, including the following. We must start small, but the growth potential is substantial.
- Core Graduate Studies: Graduate programs at other institutions, particularly outside the US, want to send students to UW-Madison to learn our core graduate sequence. This training would enhance their in-country degree programs. This may best be served by a capstone certificate with 609-610 or 709-710 and 849-850.
- Uncertainty Quantification: Teach the quantitative characterization and reduction of uncertainties in applications. Focus on methods to determine how likely certain outcomes are in partially understood systems. Focusp is on computer simulation modeling of complex systems. Develop this closely with faculty in other data science fields, notably CS and College of Engineering.
- Inference for Systems Genetics: We have several faculty who are involved in research teams at UW-Madison, and short course instruction at other institutions. We could readily develop our own Summer (Jun-Aug) or Winter (Jan) short course on this topic, likely drawing on faculty in Genetics and BMI.
- Mathematical/Statistical Finance: Core foundations (311-312 or 431) would be coupled with specialized courses in mathematical or statistical finance (to be developed). New courses could focus quickly on the study of spatio-temporal patterns (time series, GPS-like data structures) essential in financial decision making. This would ideally be done jointly with Mathematics, Economics and the School of Business.
- Big Data Analytics: This campus has tremendous leadership in big data, but there is little in the way of course offerings. We could combine local expertise with guests from industry and other institutions. We need new courses that teach how to filter big data with the aim to “throw away the chaff” and focus on the signal (data compression), how to analyze such signals, and how to visualize information at multiple scales. Emphasis will include hierarchical models and data structures. Topics and instruction will likely span statistics, computer science, and other data science fields on campus. Use and methodological development of software tools would play a major role. Develop in conjunction with other data science fields, notably Computer Science and Library Information Science.
Non-traditional Programs: These programs offer the potential of reaching new audiences and of bringing in new sources of revenue. Other institutions, notably Stanford U, have been quite successful. We would need release time and other assistance to develop these programs.
- Capstone Certificate: Develop capstone certificates for non-traditional audiences. These would likely involve existing courses aimed at novel audiences. Two natural choices are our Core Graduate Studies sequences and Mathematical/Statistical Finance.
- Professional Masters: These degrees are aimed at non-traditional audiences that are not served by any existing degree programs here. Other institutions are developing such programs, and we should be capitalizing on this opportunity. Attracting non-traditional audiences will revitalize both instruction and research, and will provide the department with novel funding sources. Investment in these programs will lead to new (or retooled) course offerings that can serve traditional students as well. Three excellent choices are Big Data Analytics, Mathematical/Statistical Finance and Uncertainty Quantification.
- Short Courses: We have several faculty who are already involved in short course instruction at other institutions. We could readily develop our own Summer (Jun-Aug) or Winter (Jan) short courses in several of the topic areas. Inference in Systems Genetics could be developed fairly quickly. I think this might be a big seller, for many researchers want to learn what they need and do not have the time to be educated broadly. We are planning to offer such short courses for our clinical translational researchers (if we can find the staff to teach).
Flexible Instruction: Educational innovation offers flexibility in modes of teaching. This will take time to develop. One important aspect of such flexibility is to enable faculty to have varied credit based on the type of teaching they undertake. Large, introductory courses should count as more than one course, while teaching an elective module (a statistical language, or an advanced graduate topic in 5-8 weeks) should count for something less than one course. In addition, advising effort should be given some form of formal instructional credit, and research experiences, including lab-style meetings, should also be recognized.