Re: Justifying Development tools

["Pat Rogers" <progers@NOclasswideSPAM.com>]

Donald Gillies wrote in message <7aa64c$m32$1@cascade.cs.ubc.ca>...
>>In article <36c85659.0@news.pacifier.com>, "Steve Doiel"
>><nospam_steved@pacifier.com> wrote:
>
>>> If you're really interested in things like the cost of errors and
the cost
>>> after shipment,
>>> read the paper by Stephen F. Zeigler, Ph.D. comparing the cost of
Ada and C
>>> development
>>>
>>>
http://www.rational.com/sitewide/support/whitepapers/dynamic.jtmpl?doc
_key=337
>
>IMHO, this article is spam science.
>
>You don't run a retroactive experiment with 7 variables and then try
>to correct for each variable independently, using ad-hoc ideas.  A
>good experiment has a "design" with just one or two variables and all
>other things are held constant.
>
>A good experiment is designed using something called "statistics"
>which seems completely unfathomable to the average software
>engineering computer scientist.  In particular, without looking at
>confidence intervals and many indepedent trials (does this guy
>understand what is the meaning of "independent trials" ???), you are
>doing nothing but promoting heresay, and we have enough newspapermen
>in the country doing this already...


What experiment are you referring to?  My recollection is that the
author makes conclusions based upon historical data.