View Single Post
  #20   Report Post  
Posted to rec.crafts.metalworking
Ed Huntress Ed Huntress is offline
external usenet poster
 
Posts: 12,529
Default Beginning programming question


"Tom Dacon" wrote in message
...

Ed, at the risk of starting a firestorm if there are any Java or C#
programmers lurking on this ng...


snip

I'm going to stop here and thank everyone for their thoughts and advice. As
I expected, these were among the most open-minded and thoughtful suggestions
I've seen anywhere on the Web. When I see similar questions asked elsewhere,
they almost always descend, as Tom suggested, into a kind of religious war.
g

For the record, my son did use Maple in college, and he has access to it
now, at work. The commercial statistics programs seem to cover the math he
needs in this work, however, so he doesn't have much need for a
Mathematica-type program now. He probably will when he starts his master's
degree program but I'm sure he'll have access to it at school.

To clarify, he does analyses of health care programs -- private, Medicare,
and Medicaid -- as a component of reports and higher analyses that go mostly
to federal agencies and Congressional committees. (This is a non-partisan
policy institute/think tank, so they're working on contracts issued by the
government and by insurance companies, not on lobbying projects.) Most of
the data he digs up is in the form of Excel spreadsheets and databases,
often with many thousands of records. He runs into the same issue that I
frequently encounter in things we discuss here, and in my article research:
Most of the data available in this world was prepared for some purpose other
than the one you have for it, so it has to be filtered, reorganized,
normalized, etc., before it can be used. Since much of it is collected by
state agencies, he often has to combine 50 different data sources into one
file, and they're all different.

He does the arithmetic parts of that in Excel. Then he imports it (usually)
into SAS, where he applies higher statistical methods. The trick to making
this work well, aside from having good ideas and insights about how to
normalize and adjust the raw data to produce the value you're looking for,
is to automate as many tasks as you can.

He's using Windows products, which often have to run all night to give him a
result. There are some minicomputers in the place running on Unix, but
they're reserved for other kinds of computational tasks.

I'm not going to pre-judge for him which way he should go with this. I'm
compiling your suggestions for his consideration. I expect that, in the end,
he'll be influenced by the programmers at work and what they encourage him
to use. But his own learning needs are a part of it, too.

Thanks again. You've all been very helpful.

--
Ed Huntress