Hi
On Sun, 5 Aug 2001 [EMAIL PROTECTED] wrote:
>
> Having taught intro stats about a dozen times now, it has not escaped my
> notice that the students do fairly well on the first two tests (primarily
> descriptive statistics) but then the first test on inferential stats, usually
> featuring z and t tests, is a massacre. Part of the problem seems to be that
> t tests are rather complicated, and very few texts give the formula for the
> test, instead taking the formula apart into standard error of the difference,
> etc. It is a confusing enterprise, IMHO.
I'm not sure what is bad about breaking t-test into SE, since
that allows logic of test to be seen (i.e., how many SEs is
observed value from expected value). I'm not sure what t-test you
start with, but the t-test for a single mean should be quite
simple generatlization from z-test for single mean. Then
progress to t for paired-differences and then independent
groups. Seems like a natural progression.
> I was thinking about changing the order of topics thusly:
>
> THE OLD ORDER:
>
> Measures of Central Tendency
> Measures of Dispersion
> Probability/Normal Curve
> Confidence Intervals
> Z tests
> T tests
> ANOVA
> Chi Square
> Regression and Correlation
I like to do binomial distribution and sign test as introduction
to hypothesis testing (before use of normal). This allows
students to actually calculate sampling distribution of test
statistic (number of successes). Then do normal approximation to
binomial, which provides nice bridge to sampling distribution for
single mean (i.e., central limit theorem). I would do confidence
intervals after the t test.
> To this:
>
> NEW ORDER
>
> Central Tendency
> Variation
> Probability/Normal Curve
> Chi-Square
> ANOVA
> Confidence Intervals
> Z-tests
> T-tests
> Regression and Correlation
This does not strike me as a good progression. Students would
jump from normal to chi-square to F back to normal (z) and then
to t. As mentioned above, I would do probability, binomial,
normal, z-test for single mean, t-test for single M, paired
difference t-test, t-test for ind groups, confidence intervals,
and then progress to the other tests (e.g., F when more than two
groups, chi^2 when both categorical variables) and to regression
and correlation.
Rather than re-ordering the material, examine closely which
transitions involve large jumps and try to break them down into
manageable steps. Although largely superfluous today, that is
why I believe instruction in binomial and normal tests provides
a good intro to inferential statistics. Also consider whether
students are getting enough exposure to the material (lectures,
labs, problems, ...), especially to applying principles to data.
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
============================================================================
