“WISE Bootstrapping Applet”
WISE Bootstrapping Applet
Nov 22, 2012
- This applet provides a “behind the scenes” view of the bootstrapping procedure. Various aspects such as population, sample size, variance, etc., may be investigated for how they affect the bootstrap resamples.
- Type of Material:
- Web applet
- Recommended Uses:
- The applet is designed to be used by an instructor, although students could play with it as well. It can be used in class or, as homework, with well-written directions.
- Technical Requirements:
- Internet browser with Java installed.
- Identify Major Learning Goals:
- The goals are: to see how bootstrapping works; to compare bootstrapping a mean and bootstrapping a median; to see the effect of population skewness; and to see the effect of sample size.
- Target Student Population:
- Students in an introductory course who are seeing bootstrapping for the first time.
- Prerequisite Knowledge or Skills:
- Students need to have a solid conceptual knowledge about sampling distributions and descriptive statistics (mean and variance/std). It would also be helpful if students are familiar with confidence intervals and with the idea of what bootstrapping is.
- It is more flexible than other bootstrapping applets because the population being sampled from is completely determined by the user. It accurately portrays the process of bootstrapping data and the directions provide good cues for exploratory learning. The applet uses good graphics to illustrate ideas.
- Because the only preset population is the normal distribution, it may be difficult to use in a lab setting if you want everyone to get similar results in the end (if you want a non-normal population). It would be nice if the confidence interval that was created was shown outside of the graphical display, perhaps somewhere off to the side. Also, bootstrapping is a very general technique, with only two uses demonstrated here.
- This can be used to display step-by-step how the bootstrap builds a confidence interval. It can also be used to quickly pull an entire sampling distribution from a bootstrapped sample. It is also nice to be able to show students how sample size affects the sampling distribution. The applet should help students understand the concept of bootstrapping. The screen showing points falling from the population into the sample and then from the sample into a resample is good.
- Without instructor guidance some of the big ideas may be lost on the students. If assignments were written for this, they would need to be very detailed or else many students would not comprehend the major learning objectives.
Only means and medians are shown, rather than more interesting statistics for which bootstrapping may be more clearly needed. I don't know an easy way to assess student learning of this content. Also, no assessment activities are suggested.
- The design is great and the visuals are better than most bootstrapping applets out there. It allows the user to see the details of one bootstrap resample or many (10,000). This applet is easy to use without even reading the instructions. It is easy to change the sample size.
- The instructions are well written, but lengthy to read through in a class setting. I would opt for orally explaining the applet to my students.
I found it cumbersome to alter the shape of the population. I also wanted to be able to choose a number of bootstrap samples other than 1 or 10,000. It would be nice to be able to choose confidence levels other than 95%, two-sided
- Other Issues and Comments:
- Overall, this applet is nice but is limited in what it presents: bootstrapping a mean or a median. Still, it would be superb for lecture or lab explorations, and it's a great applet that can be used to illustrate how a bootstrap sample produces a sampling distribution and confidence interval.
- Creative Commons: