Hi
Many Psych students do need lots of assistance to have a
successful experience in stats courses. Some of the things that
I have tried over the years include:
1. Give them problems to work on outside of class (e.g.,
assignments that might not be worth too many marks) and encourage
them to work in groups. For my honours class I use quite
elaborate assignments that take much time, but allocate few
grades. Forces students to essentially "study" (i.e., learn the
material in order to do the assignment).
2. Use lots of literal repetition. Do not assume that because
you have done it once in class or they were suppose to have
learned it in a previous class (and probably did) that they will
remember it a month later (or less for some). I am shameless
about saying the exact same thing repeatedly, and will even joke
about it. A number of years ago, I used an
accelerated/repetition approach to intro stats. It was a full
year course. Each term we went through all the material in about
2/3rds of the term and had a test. Students who did well had the
option of tutoring other students rather than doing later quizzes
as we went through the material a second time in an abbreviated
manner. There was some suggestion that the method was effective
(it was a multi-section course with common Xmas and Spring
tests) and it makes sense that it would. Students often complain
about stats that it made sense only after the course was
over. Rigorous assignments/quizzes that require students to work
along during the term probably achieve something of the same
benefit.
3. Be sure to focus on the principles, wherever possible using
conceptual repetition. That is, find alternative ways of
communicating the meaning of the constructs. I teach 3-4 ways of
thinking about difficult concepts like part correlation or
interaction. The computer can be an effective tool in this
regard (e.g., generating residual predictor scores to correlate
with the dep var to obtain the part r), but should not be allowed
to become too central. Using the computer is a fairly trivial
task (although I appreciate how challenging some students find
it).
4. Carefully coordinate all elements of the course to converge
on learning the important material. I use my own manuscript
(which has alternating conceptual and SPSS chapters), lecture
largely from the text using overheads of tables and figures
(which students have because they are encouraged to bring the
text to class), have paired labs in which students alternately
perform manual calculations using just formula sheets and perform
same analyses on computer, assignments involve generating
simulated data and doing the calculation/explanation/SPSS
analyses presented in class, and tests evaluate in the same
manner (i.e., data, SPSS printouts, formula sheet). The
essential aim is to get students to perform calculations and
provide explanations that demonstrate their understanding of the
material, and to relate that work to SPSS printouts. The
calculations are essential parts of understanding the material
(e.g., showing that part r on printout equals change in SS
predicted with addition of that predictor over the SS total).
5. Help students to develop a sense of the cumulative nature of
the knowledge that they are acquiring. I start this from the
very beginning by showing how what seem like complex formula
(e.g., that for r) actually are quite simple if you think of them
in terms of components (i.e., SCP, SSx, SSy), rather than as
arithmetic operations. In class and labs, and on assignments and
tests, I also emphasize that students need to demonstrate the new
knowledge at each stage (e.g., explaining SS and SD is important
first time, but not later; calculating r is necessary for first
simple corr/reg problem but not later when simple rs are being
used to calculate MR regression coefficients).
6. Try to develop general schemas that help students to
appreciate the commonalities across superficially different
statistics. For example, T-test can be conceptualized as some
parameter (p) minus some hypothesized value for the parameter (P,
often 0), divided by its standard error. That is, t =
(p-P)/SEp. Students then need to learn different SEs for single
sample mean, difference between paired scores (a variant of
single sample), difference between independent observations, r,
regression coefficients (simple and multiple), ...
7. Build in strong sequential dependencies, emphasizing early on
things that will be important later. For example, it helps when
understanding regression to use the (n-1)s^2 formula to obtain
SSs (e.g., for total, predicted, residual). So earlier when
teaching SS, SD, and s^2, put some emphasis on the fact that s^s
= SS/(n-1) means that SS = (n-1)s^2. When teaching simple
regression, point out repeatedly that residual Y scores are
uncorrelated with X (e.g., by having them generate correlation
matrix in SPSS), so that later this can be used to help
understand part correlation. You need to ask yourself more than
with most courses what prior knowledge do students need to
understand and be fluent with to grasp the current material.
8. Help students develop a decision-tree for deciding what stats
are appropriate for different designs. I tend to do this in a
piece-meal fashion, adding as we go along in the course. In
general I tend to prefer revealing the structure of things in a
gradual manner, rather than having an overly regimented story set
out from the beginning. I worry that students will simply
"memorize" the structure, rather than working at understanding
and building it themselves.
9. Develop the skill of answering questions in a manner that
does not reveal more than is necessary. That is, try giving
prompts and hints before stating the answer outright. Students
do get quite frustrated with this approach at times, so again
try to be somewhat light about yourself.
10. Be prepared to help with some pretty basic things, such as
the number facts mentioned by others. Another example of this is
the use of calculators. Many students will not know that their
calculator has a memory, let alone how to use it. And you would
be surprised how many different ways there are to design a
calculator keypad so that one can store, sum, and retrieve
numbers! Even after many years, I come across variations that I
cannot figure out for the life of me.
11. Practice making mistakes gracefully. It is inevitable in
stats that you will make some error in front of the class,
especially if you occasionally go over material quite quickly
(e.g., when repeating or reviewing). Happily students are
generally quite forgiving.
12. Get in the habit of thinking out loud as you lecture. Try
to verbalize even the most mundane and simplistic of steps.
13. Monitor class comprehension as best as possible. Look for
the quizzical look, and encourage questions and participation
from the class. When in doubt, err on the side of repetition and
further simplification.
14. Consider carefully where your course falls in the grand
scheme of things. It is impossible to teach anything well in a
short course if you try to teach too much, and it is simply
impossible to include all of the things that could (not should)
be taught about stats.
Although challenging, teaching stats can be most rewarding, as
even good students often feel (wrongly) that they just can't do
this kind of work. It is nice to see them learn otherwise.
Best wishes
Jim
============================================================================
James M. Clark (204) 786-9757
Department of Psychology (204) 774-4134 Fax
University of Winnipeg 4L05D
Winnipeg, Manitoba R3B 2E9 [EMAIL PROTECTED]
CANADA http://www.uwinnipeg.ca/~clark
============================================================================
