Peter Flom writes:
I am now up for a position which would require teaching some introductory statistics to people studying to work in health care. Mostly, these people will have only a HS diploma, and it may be a fairly old HS diploma (a lot of them are returning to school).
For the interview, though, I am assigned to give a 30 minute talk (no powerpoint or anything, just a white board).
I am told that:
You can assume that they are familiar with basic descriptive statistics such as means, medians and standard deviations, and that they understand the distinction between a sample and a population. Your job is to explain, perhaps illustrating with an example, how to use sample data to make inferences concerning the difference between two means. Naturally, we don’t expect you to treat this subject exhaustively in half an hour, but you should attempt to make a semi-complete “module” out of the lecture, rather than simply stopping part-way through the material.
I’ve just started thinking about how to do this; I’m going to look through some of my elementary stats books. But if you have any advice, or a book or website to suggest, I would appreciate it.
My reply:
I’m really no expert here, but here are my quick thoughts, which you should think of as suggestions rather than a plan to follow.
– Start by telling the students what they should be able to do at the end of the lecture that they can’t already do.
– Start with an applied example–something relating to health care that’s newsworthy–a story which you can then abstract into a simple numerical example.
– Then do your procedure on the numerical example. I assume you’ll be computing the mean of each group, subtract, hen use the formula to get the standard error for the comparison, and then get a confidence interval and make a statement about statistical significance.
– Explain what the procedure does, how it works. Don’t be afraid to frankly treat it as a cookbook formula. Cookbook formulas are good. That’s what we all use for things that we’re not experts on. The goal is for the students to be able to take the cookbook and understand how to modify it.
– Give another numerical example. I’d suggest the whiteboard, but you don’t know how much whiteboard space you’ll have, so to be on the safe side, bring this one as a handout. Have students work on it in pairs–give them 3 minutes. It shouldn’t be too hard, because they should be able to follow the template on the board. (If there are no “real” students in the room, just evaluators, do this anyway–they’ll play along with you.)
– Go back to the applied example and explain the relevance of the statistical method. Tell a story from your own applied experience.
– You should have 5 minutes left. Talk about the limitations of the procedure and how this connects to future course material. If I were giving this part of the lecture, I’d focus on issues such as potential lurking variables and the use of regression to control for differences between the groups. Some people would yammer on about normality and heteroscedasticity, but, as you know, I think that sort of thing is (a) a distraction and (b) leads students toward technical issues rather than toward the substance (heath care) which is where they can more usefully direct their thoughts.
– Conclude by telling the students what they can now do that they couldn’t do before.
Whiteboard technique: Treat the whiteboard not as an extension of your short-term memory but rather as a three-dimensional document (with the three dimensions being horizontal, vertical, and time). When all is done, the whiteboard should be readable and coherent. You’re preparing this for a job interview, so you should definitely “cheat” by preparing your lecture ahead of time so you’ll know what to put where. (If the whiteboard is small, you can fill it up, erase it, and fill it up again. The point is for it to be coherent, with readable headings for each part.)
Good luck!
P.S. Yes, I would recommend doing one of the class participation demos from our Teaching Statistics book: we have a chapter on statistical inference, you could use one of these.
It's too bad you cannot choose your own topic. Perhaps you could ask for an additional 15 minutes to do the illustration of Fisher's tea tasting exercise as described in Teaching Statistics as an example of an experiment. This really gets the students involved. I've added one twist to this where I do it with a second student but insist that the second student cannot do any tasting, i.e., they must randomly pick four cups as their selections. Once I had the first student selec the correct four and the second student (without tasting) also do the same. For more advanced audiences perhaps one could discuss the notion of whether the taster assumes that the cups are arranged in blocks of two (see Senn's Dicing with Death for a nice discussion of this). good luck.
30 minutes is too short to do anything useful.
Thanks, Andrew!
Probably you know this presentation from Hans Rosling who accomplished the desired result in 20 minutes.
He did it with some passion and a visual tool that shows highly visual mean shifts over time see http://www.ted.com/talks/hans_rosling_shows_the_b…
What seems to missing is a pedagogy that helps people put abstractions into concrete use and communicate them. It is the learning dilemma of future pilots who begin as novices and a manual of abstractions about flying but have no flying experience. Advanced beginners have read the manual, maybe know that when you look out the airplane window blue is up and green is down. Pilots accumulate experience. Experts improve flying skills. Masters land planes in completely new situations like the Hudson River.
I used to have to teach engineers statistical process control in the factory. Straight out of university they all had "a great set of class notes" but no experience using mean or range shift from data from the xyz machine and knowing how to use it to reduce product cost or improve quality. My challenge to them was to put their abstractions to use and communicate improvement results to others.
I think there is a pedagogical blind spot here, we seem to loose our way in our own abstractions. To paraphrase the learning theorist Piaget; learning involves finding abstract principles in concrete experience, using abstract ideas in real situations, taking cases in one domain and putting them to work in another.
Recently I had the task of teaching some young native Spanish speaker ecological scientists to write their cases skillfully in English language technical journals. It was a Yogi Berra deja vu moment all over again.
They had the concrete experiences, They could collect data and statistical models with SPSS and the rest; but could they make their case insights available to others? Tried to read any technical journals in Ecology lately? Wow.
Practice, practice several time yuour speech.
Given the described background of the students, I'm not sure I'd assume that they "understand" standard deviations, or even medians. I think it couldn't hurt to review this to make sure there is really no question. If they don't understand standard deviation or how it's useful in statistics, they're going to be following the cookbook formulas without even a basic understanding of what the formulas do.
I have taught coming-back-to-school nurses often. They always do better with the concrete example first, then the abstraction from it later. Their grasp of the concept is always enhanced by graphs. The oft-maligned box-plots make it very easy for them to grasp both differences in central tendency and differences in variability, as a set-up to the hypothesis test. They are often intrigued with data that upset existing paradigms (e.g., the heart attack rate in women taking hormone replacement therapy vs those without – when the prevailing paradigm was that hormone therapy protected against heart disease.)
Good luck!
Thanks to all for the helpful suggestions.