"Bob LeChevalier" <lojbab at lojban.org> wrote in message
news:1k37juss6m9iuahd47ubpglru223okfpj2 at 4ax.com...
> "Parse Tree" <parsetree at hotmail.com> wrote:
> >A properly constructed test wouldn't have been multiple choice. The test
is
> >essentially meaningless.
> >
> >I like how you admit that they're multiple choice now.
>> Actually the test is well-constructed, since the various multiple choice
> answers each corresponds to a particular conception or likely
misconception
> of the problem. When one is constructing a normed test, guessing is not a
> problem. One can adjust scores for guessing by including a guessing
penalty
> (this happens on the SATs for example) or one can simply compare
percentages
> choosing each answer with those of other populations (which is what TIMSS
> does).
The SATs are not well constructed. Generally, guessing penalties don't
work, and there are numerous reasons for this. Firstly, guessing penalties
rely on a completely random selection, which is rarely the case. Actual
selection is dependent on what they've been taught and how they work it out
on their own. You know, many teachers teach things incorrectly, because
they are not qualified in the subject they are lecturing on.
> TIMSS is not designed so that ANY kid could get a perfect score or
> anywhere near a perfect score, and I've never read a report that suggests
> that any kid did so. There are easy problems and there are hard problems,
> and the problems are weighted by the difficulty that they were found to
> present to the entire test population.
That's probably because children who answered every question correctly were
penalized for not showing enough work, or showing the work that the markers
wished to be shown.
> TIMSS also included many problems that were NOT multiple choice, BTW, and
you
> could not get full credit unless your work was shown and contained the key
> steps expected in the solution.
This is also bad. While better than multiple choice, it still allows a
large quantity of discretion. In many disciplines there are numerous ways
to come up with a solution to a problem. Statistics itself has this.
Also, key steps implies a great deal of cultural and related bias. The
amount of work shown by someone confident and familiar with type of question
is quite a bit different from that shown by someone not confident.
Seems to be quite a bit of memorization for these tests.
