Re: Potential new project: Calculator (scientific, graphing,statistical)

["Pete St. Onge" <pete@seul.org>]

Bill Tihen wrote:
> 
> I have wished for such a program for a while.  I found the current programs
> too programming oriented for a good math lesson.  But could you please
> model it after the sharp graphic calculators -- not the TI-series?
> 
> The statistics part is an excellent idea.  I would also really like it if
> it had stats that allowed you to some best fits & stats for physics classes
> -- like error bars to see how well the equation fits the data.
> 



> A diversion --
> (I always thought it was odd that when 20 kids are doing the same
> experiment (with the same procedure) that they don't collect the
> data and do some real analysis like I did in as a researcher.  It
> seems so logical especially with the new NCTM standards that many
> schools are working towards.  (However, I guess that is the math
> dept. not the science dept.  I also know that biologists do a
> different sort of statistical analysis (although not being a biology
> researcher I can't say what they do with the data).
     From what I've seen thus far, most of the tests I see used commonly
for biological data are t-tests, F-tests, regression (OLS), chi-square.
Some graphing would be useful - GUPPI can be used for much of that.

<MODE="WARNING: Low sleep, high caffeine">
     I've been thinking of what is really needed in a good, basic data
analysis package ... a spreadsheet-like data page with links to
databases and the ability to add analysis 'modules' so that you can add
more functionality as you need it without the beast becoming cumbersome.
A data-focused approach to analysis rather than a strictly
procedure-based approach that most analysis packages take (ie. SAS,
Systat, SPSS). Graphical data exploration using the OpenGL libraries and
a mouse or joystick. Means to test the validity of the assumptions
behind analyses (normal distribution, A transcript of all steps taken
during analysis so that one might go back and try different things. The
means for more than one machine can connect to the analysis in progress,
being able to watch and contribute to the analysis.
     This is a tall order, I know. Having a package that can scale up
from something usable by high school and undergrad students to something
that can be pushed and pushed hard without crashing by more robust
analyses would allow a tremendous 'user base' to build up
     
> Anyway, in my oppinion this would be valuable if it could be used to
> help collect data from various sources.  Help find equations and do
> the anaylsis needed for both math and science classes.  And with the
> ease of a graphing calculator.
     Being able to connect to text files (making it simple to bring in a
lot of data ... students can build simple files easily), other
spreadsheet formats, database backends, etc, would make it very
versatile. Being able to deal with different types of equations,
analyses, regression types, etc., would make it very popular and would
very likely the 'killer app' for math and sciences.
     Make the interface pretty simple, few buttons, mostly context menus
so that the user can literally explore the data space.

     This brings up the point about data management, particularly for
those involved with more complex data. The hardcore SQL approach to
database analysis is inefficient when it comes to dealing with data
having structure (spatiotemporal). I can illustrate with an example in
my own field. My physical limnology work often requires that I calculate
lake volume for all or part of a lake. If the 'lake' is divided into 0.5
m sections for the purposes of measurement, and each lake is different
with respect to depth, it becomes difficult to use a summary-type SQL
query to get lake volume.
     The case could easily be made for one to set up the table in
question not having depth and planar surface area at depth (the info
most used to calculate volume), but rather as the results of said
calculations (ie. lake volume calculated from the given depth to the
bottom of the lake, for instance). The point is moot as I also need the
planar surface area for other calculations. :)

     Alternatively, Spreadsheets are very flexible, to the point where
they don't directly have the means to validate and verify data (Remember
"GIGO"?). They also tend to be one of the most intuitive packages for
analysis ... few researchers or math students don't know how to use a
spreadsheet. As flexible and useful as it is, spreadsheets aren't great
for large, complex analyses.
     What is needed is something which can bridge the gap between the
two. Though I've done some work to this effect (ie. LimnoTools -
http://www.trentu.ca/~erpds/PLT1.html ) and it has been very useful in
my own work, the implementation language (VBA) is a bit ungainly for the
performance such tools would have to have.

So, to summarize, a good data exploration and analysis app would need to
be:

- scalable: > not too big for small needs, not too weak for great needs
- flexible: > ability to carry out individual fine calculations &
              manipulations
- extensible: able to add new graph and analysis types, and build new
              types from existing types
- usable:   > adaptable to different tasks, flexibility to accept
              data from many sources
- intuitive > graphical representation / participation paradigm rather
              than process-based paradigm
            > good docs
- auditable > able to see previous analyses

</MODE>

Anyhow, my two cents or so.

Cheers,

Pete

-- 
Pete St. Onge - McGill U.  Limnology - Fun with Ropes & Buckets
pete@seul.org                  http://wwp.mirabilis.com/4322052
---------------------------------------------------------------
SEUL - Linux for All!                       http://www.seul.org
Programming For Science Page        http://www.trentu.ca/~erpds