There is not a straightforward way to do a Student t-test in Ferret,
but such a script could be written, I think. It requires evaluating
the Gamma function, which would be tricky in Ferret because it is an
integral that is evaluated until it converges. I've attached the
fortran code I use (from Numerical Recipes) in case someone smarter
than me wants to write it into a Ferret script (and please post to the
list!).
The Wikipedia article (Student_t_distribution) is pretty good.
One very important thing to note when doing a Student t-test is that
you must know the degrees of freedom first! That is a separate
calculation that is typically found as an integral over the product of
the autocorrelation functions of the two variables. There are several
other essential subtleties, that are described in, among other places,
p3015 of Kessler et al (1996, JPO, link to the pdf at the end of this
message).
Statistics are hard.
Billy K
Link to the paper with description of the procedure and assumptions
(pdf, see p3015):
http://faculty.washington.edu/kessler/abstracts/ksmh-1996.pdf
On 21 May 2010, at 9:12 AM, Steve Hankin wrote:
Hi Sarah,
If you have a variable that contains the statistical confidence
values, then the result is easy:
LET var95 = IF confidence GE 0.95 THEN my_var
The values in var95 will be masked to include only those where you
have 95% confidence. Others will be set to "missing" and become
invisible.
I am not aware of a student t-test function in Ferret, but it is the
sort of calculation that can be readily put into an external function
(and thereafter made sharable for other Ferret users).
- Steve
=======================================
Sarah Bonham wrote:
Dear Ferret users,
I have 2 sets of temperature data, both on a 73x96 array and I am
trying to produce a map of the difference between the two showing
only temperature values that have a 95% confidence.
Is there a function in ferret to apply a student t-test or something
similar?
Ideally, I would just like to map the temperatures that have 95%
confidence, but if this isn't possible and there is a way to put
shading over the top instead, then this would be great to.
Any help is greatly appreciated,
thank you in advance,
Sarah
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
William S. Kessler
NOAA / Pacific Marine Environmental Laboratory
7600 Sand Point Way NE
Seattle WA 98115 USA
william.s.kessler@xxxxxxxx
Tel: 206-526-6221
Fax: 206-526-6744
Web: http://www.pmel.noaa.gov/people/kessler