Dear Ron,
 
I taught biostatistics in Public Health for about 15 years. With some extra planning effort, there is no other way to go.
 
I slowly decreased the number textbook problems and increased the number of laboratory type exercises over time.
 
For example, to introduce the software I did the following. I found a data set I liked, boating accidents over time in the US, reported by the Coast Guard, and gave the students my own analysis of quarterly data using graphs, tables, etc. All purely descriptive. I then e-mailed the materials and data or put them on Blackboard, then put them in the  computer room, and got them started doing their own analysis to reproduce the report but using a different time interval for everything. The group part was to plan their analysis, even though they had a prototype.  When they asked how to do a specific thing my usual response was to help them find the page in their software manual.
 
I always chose a software package that could be learned in 30-60 minutes. I preferred Stata. SAS is clearly a dog and should be avoided like the H1N1 or worse.
 
Each week after the RA, I had groups work together to plan their textbook problem solutions and then give them to the whole class. Each group had 2-5 problems.
 
I never succeeded in having them plan a problem solution and then assess whether they would get a solution. They had been severely brainwashed by American math education into believing that each problem had one solution; they were determined to find it for each problem sequentially.
 
Real problems in statistics don't have clear, simple solutions. Everything in real statistics is about exercising thoughtful judgement: choosing the p-value, deciding if the data satisfies the usual assumptions, choosing a test among plausible alternatives, and deciding how to write about it.
 
The only way I could overcome this is to give them problems that that complex solutions, such as the boating accidents, or different solutions, eg, each student had a different data set, and required preparing a more sophisticated answer. After a few years about half their work was on textbook problems (my friend Andy described these as "toy problems")  and about half was on more complex cases that I had designed myself.
 
If you have "writing across the curriculum" like most places now, you can have some good writing assignments with statistics. For example, they can rewrite something that has been published but to a specific standard for information and clarity. People have to explain what they have done and what it means, to different audiences. There is some good material today on clear use of tables and graphics with data. Choosing to use a graph, table, or text is a challenge.
 
You can satisfy most of the student demands for foolish talk with a mini-lecture on each topic.
 
Most statistical topics take a week in an introductory course. It is nearly impossible to ask them to read two or three chapters for an RA. (They know this is just wrong!) I ended up giving an RA nearly every week as a result. A text that organizes statistical methods in a different way might facilitate more integrated presentation of several topics but the books just aren't written that way.
 
I used TBL in courses that met once a week for three hours and twice a week for an hour and a half. The rhythm of events was different but both could be worked out. For the former I had to go from an RA through group presentations and on to a mini-lecture on the next topic all in one session. For the latter the group work on problems always broke over two sessions. The latter was more flexible since I could schedule some topics for one session and others for two or occasionally three.
 
I stopped giving the students autonomy in the grading components after about three years. No one appreciated it, coming to agreement was nearly impossible, and they thought they could strongly manipulate their grades through it.
 
I rarely had groups that worked poorly. I used groups of 5 or 6, never 7, and rarely 4. I know what others recommend about smaller groups but my least successful groups were always size 4.
 
But don't ever forget that they have brainwashed about what math is. They believe that statistics is just the same. Some students will be very unhappy about not finding the same thing in your course. But they can't if you use TBL.
 
Best wishes,
 
David Smith
 
On Wed, Jun 23, 2010 at 2:43 PM, Ronald Giachetti <[log in to unmask]> wrote:
Hi All,

In the fall semester I'll be teaching a course titled, "evaluation of engineering data" that involves probability and statistics.  I want to try TBL for this course, and I'm trying to think of good team assignments for in the classroom.  The problem I find is that the course material is usually presented as black & white -- in other words, for a problem there is only one correct answer.  In all the courses that I've seen that use TBL, the team assignments are such that there are multiple good answers and arriving at an answer requires a team to discuss and negotiate the strengths/weaknesses of different answers.

I've developed classroom exercises to demonstrate ideas.  For example, for confidence intervals I have a bag full of numbers from a distribution.  I have each team pull a sample of numbers and construct a confidence interval.  Each team then draws their confidence interval on the board.  I then draw the actual population mean and show them that even though all their confidence intervals are different, they all contain the mean.

While I think the above exercise is better than straight lecturing, it doesn't capture the team-based learning that I've seen in other classes.  So, if anybody has any experience in quantitative courses or can direct me to literature on the area, I'll be interested in how they construct team exercises.

Thanks.

Ron.




Ronald Giachetti
Associate Professor
Department of Industrial & Systems Engineering
Florida International University
http://web.eng.fiu.edu/ronald/
[log in to unmask]
305-348-2980



--
David W. Smith, Ph.D., MPH
Chartered Statistician