Webinars

We will discuss how we coordinated, held, and judged a wine pricing competition (hosted on Kaggle-in-Class - inclass.kaggle.com) to engage students in applying prediction techniques learned in our data mining class at Stanford. We found that with proper incentives, the competition was very successful in getting students interested in working collaboratively in a race against the clock to eke out additional predictive performance in their models.

The role that data collection plays in causal inference is of fundamental importance in introductory statistics, and yet is outside the comfort zone for many of us. In this webinar, I'll discuss why causal inference is important and also fun, and give some advice for teaching this topic.

A challenge in introductory statistics is to motivate the estimation of unknown population parameters. In this activity, we allow students to estimate the proportion of the continental United States that is within a mile of a road by repeatedly sampling latitudes and longitudes and viewing that location using an internet mapping service. Technology is used to generate random values within a specified geographic rectangle, which populate a data collection spreadsheet. Students are instructed how to use MapQuest.com to determine if the random location is within the continental US and if so, whether it is within a mile of a road. This data collection task helps to fix ideas of study design ("what if the point lands in the middle of one of the Great Lakes"?) as well as motivate the estimation of an unknown proportion. Individual confidence intervals can be created and compared, as well as creation of a class-wide confidence interval. This activity can be used in introductory classes at all levels.

Generally in learning statistical inference, students reason backwards from data to the (usually invisible) process that produced them. I will demonstrate an alternative approach in which students begin at the process end, designing their own "data factories." Based on their output, students modify their factories such that, for example, a collection of cats produced by a cat factory has features that look more like real cats. This work is part of the NSF-funded "Model Chance" project. In this project, we have been adding probability modeling to the existing data-visualization capabilities of TinkerPlots and, using that environment, exploring how data and chance might be better integrated in our instruction beginning in the middle school.

Formative assessment is where feedback on learning activities is used to modify the method of teaching to meet the needs of the learner. One such strategy is the use of personal response systems, or clickers, for instant feedback. Immediately examining the responses, teachers transcend the lecture-only model and are empowered to foster student-centered learning by explaining misconceptions or feeling confident that students understand the concepts. Attention is no longer deferred to the loudest student or the fastest hand-raiser, but rather to entire class. I have wanted to try clickers, but was reluctant due to the barriers of implementation - cost to the student and software and hardware demands, including students forgetting their clickers. I recently discovered polleverywhere.com, which allows students to text in their answers using cell phones and see the results immediately on the screen. Results can be captured and shared with students on websites or blogs. It's free for small classes and claims to cost 1/3 as much as clickers for larger classes. My students love it! I'll discuss my experiences and share how I have integrated some of the classic active-based exercises in statistics into my class and used formative assessment techniques that have helped bring my classroom to life. Have your cell phones ready if you join this webinar!

Developmental education in America's community colleges has been a burial ground for the aspirations of our students seeking to improve their lives through education. Under the leadership for the Carnegie Foundation for the Advancement of Teaching and the Charles A. Dana Center, nineteen community colleges and systems are building accelerated pathways to and through developmental education with the goal of helping students with low levels of mathematical preparation complete a college credit bearing, transferable statistics course within one year. Uri will describe the work to date, the challenges the initiative faces, and the underlying ideas of improvement science that are driving its development.

One challenge in any introductory statistics course is helping our students understand the logic of hypothesis testing. In this webinar I'll demonstrate one of my favorite examples for doing this. The data are a sample of golfballs. The hypothesis is that the number on a golfball is equally like to be a 1, 2, 3, or 4. Using a function written in R, I allow students to design their own test statistics and then produce a graphical display of the sampling distribution and calculate empirical p-values. This activity can be used in introductory classes at all levels - even if you don't cover goodness-of-fit testing. It can be used as a first introduction to inference, as a motivation for the chi-squared test statistic, as an example of goodness of fit testing, or as a demonstration of simulation-based inference.

Since formal hypothesis testing and inference methods can be a challenging topic for students to tackle, introducing informal inference early in a course is a useful way of helping students understand the concept of a null distribution and how to make decisions about whether to reject it. We will present two computer labs, both using Fathom, that illustrate these concepts using permutation in a setting where students will be answering interesting investigative questions with real data.

When instructors have their students implement "real-world" projects in statistics, a number of questions arise: Where can students locate real data to analyze? What kinds of meaningful research questions can we help students to formulate? What aspects of statistical research can be covered in a project? What are reasonable methods for evaluating the student's work? The presenters will share resources developed during an NSF-funded study to develop and test curriculum materials for student projects in statistics, using linear regression and t-test scenarios.

Some students in statistics classes exhibit behaviors that share characteristics with the established construct of learned helplessness. This webinar will discuss this phenomenon, and detail an instrument recently developed which measures this (HILS: Helplessness in Learning Statistics).