Hardly a summary: I had two responses. But in case you are interested:
I had posted:
"I was impressed with Andras Vag's comprehensive reply just now, but
wondered: how do you know that the online calculator is giving you the right
answer ? In the pharmaceutical sector, "everything" has to be validated,
including software. If one uses the online services, is it at one's own
risk? ~
is it advisable to first validate it yourself (for the use one is making
of it ?). Has anyone had a bad experience with one of these web-based tools,
which, if valid, would be very useful ?"
Interestingly, noone replied quoting a bad experience.
Jim Groenveld replied with:
"Hi Alan and Martin,
See also the paragraph "Stat Calc Web" at
http://home.hccnet.nl/jim.groeneveld/links/Statist.shtml
When I use these calculators (which may differ quite a lot in appearance
(GUI), support, statistics, etc.) I always use more of them for the same
calculations. I do this not only to check them, but also to check myself,
whether I selected the right calculator / test. Results, especially sample
sizes and the like may differ a little bit. Sometimes, if possible, I also
check with PASS, SAS, Minitab and S-plus, and an internal statistics package
within Medtronic: MedStat. The advantage of some of the on-line and off-line
calculators are that you can quickly perform all kinds of scenarios."
Tobias Verbeke replied with:
"One general comment. If one uses open source software, one can be
absolutely
sure that the right thing is done, because the source code of every single
data-analytic procedure can be read and verified. Even for the
pharmaceutical
sector this cannot be said, at least about SAS which is heavily used in
clinical
research.
One of the web-based solutions cited by a colleague (Rweb), uses the
open source
environment for data analysis R (www.r-project.org) as a backend. The
highly competent
development core team and large user base of knowledgeable statistical
programmers makes R a *very* reliable tool, which gives access to the
latest developments in statistical analysis. It has indeed more or less
become the lingua franca of theoretical statisticians.
As you can understand, I rely on R on a daily basis. The government
agency I work
for requires the level of transparency and quality control it offers. R
is, by the way,
not only free as in free beer, but also as in free speech.. (www.fsf.org)."
Best Regards,
Martin
|