Re: [school-discuss] Collecting Assessments at District Level

[mikee <meschman@engima.com>]

hello,

we are helping (i.e. bidding at ) several school districts deal with 
assessment data.  currently, this district has teachers prepare excel 
spreadsheets and access databases that reflect the status of individual 
classes and schools.  the baseline data is consistent in content, but not in 
format.  that is - cell positions are not standardized.

we have suggested the following :

1 - the existing templates will be uploaded to a postgres database.
postgres allows multiple logical views of the data with a flexibility not
possible in mysql.  mysql is, however, preferable, because it has a large 
stable user base and multiple porting tools for both access and excel.
thankfully, data migration between mysql and postgres is trivial in
most cases.

2 - in cron functions, batch procedures will migrate the data to a 
universal template and rollups will be performed,  the results being placed 
in a mysql results grid which can generate pie charts, bar charts, x-y graphs,
and 3d topologies.  the 3d topologies provide a visual means for an assessor 
to measure the variance between expected results and actual results.

we have introduced this feature (3D topological assessment of data in a time 
series animation) without charge, to induce our customer to consider its use.
we have Been doing this sort of thing for five years to measure coastal 
erosion, weather event effects on drainage systems and seafood crop cycles.

these variance data are used by assessors as one measure of the validity and
utility of a given data mapping. (now with students instead of rainfall or 
oysters).

3 - we feel it is essential to cross-index all data as richly as possible, 
since assessors plans to index data presume that a given ruler will produce 
the measurements necessary to both prove compliance and to formulate
changes to operating procedures.  

because of the expense of creating software to index results,  and because 
changes in operating procedures, general conditions of the community and
unforeseen factors are part of the circumstance, a scheme of measurement can 
invalidate itself :

that is, the negative factor has been neutralized so the indexing results are 
no longer interesting.  or, the compliance metric is altered so the index is 
meaningless.

4 - batch procedures running on cron are cheap and take place off hours,
i.e. they do not affect non-incremental costs.  on-line access to results are 
not impeded and what-if scenarios are more easily supported.

5 - the reason we place such an emphasis on the service level is to encourage
competition through demonstration among the potential assessors.  we want the 
data delivery mechanism to encourage assessors to postulate conclusions, then 
verify them.    assessors can make their cases on their own terms and still 
provide statistics on a uniform metric used for compliance reporting.

6 - so only one (albeit large and complex) data template is used for 
compliance reporting, but every data analysis scheme with promise can be 
explored without additional non-incremental costs.  every assessor has a 
reason for preferring a particular data scheme and this design principle 
allows those reasons to be followed to a logical conclusion.

note : remember we are only talking about dumping matrices to lists,
resorting and / or condensing them, then reloading the results into 
other matrices.

the amount and volatility of assessment data, and the access patterns of that 
data are not significant when compared to i/o access for an administrative 
application such as payroll, so it makes sense to slice it every way 
imaginable.

7 - when we are finished, teachers will continue to use their existing, 
non-uniform templates (they already know how to use them) and uniform results 
(and alternate views) are available on-line.  if a mis-step in data analysis 
causes the district administration to recast data assessment and reporting,
impact on teachers time is minimized.

8 - we hope this will encourage users to explore the data, and to begin 
reaching conclusions by following hunches and modeling them.  users can gain 
insight from the data then track actuals against the insightful models.  if a 
clear trend emerges, they have a non-social, non-political venue   in which
those insights can be presented.

9 - no one in any field can predict the veins of meaning a large database can 
contain.  these veins of meaning must be discovered.  in that light,
every indexing proposal has some merit.  with our procedural base,
these avenues can be explored as a part of the routine, and merit in an 
indexing scheme loses its personality.  pandemonium is averted, because 
petitioners have to adhere to quotas.  

"you've had your three models incorporated and that's what we agreed to.  it 
will take additional funds to  create a fourth, unless you use part of 
someone else's allotment."

10 - thinking this way is the natural outcome of placing indexing of data in 
a service level context.  to find a winning horse, lots of small bets must be 
placed.

11 - using topological maps to find subtle variance over time between 
anticipated and actual results is a powerful technique used in simulation 
environments.  i hope it gets explored in this schools assessment area, 
because the area is so important.

we can discuss this further if interests dictate,  offline or in the public 
forum where i displayed an inadvertent lack of manners.

mike eschman, etc...
"Not just an afterthought  







On Thursday 14 March 2002 07:14 am, you wrote:
> Mike,
> That's interesting hardware information (particularly the WalMart
> machines ;^) ).  I actually never really thought through the basics of
> data in relation to hardware.  I'll keep those ideas in mind.