computing seasonal variances

[Helene Brogniez <hbrogniez@xxxxxxxxxxxxxxxxxxx>]

Dear Ferreters,

I am working on 14 years of monthly data of observed tropospheric 
humidity and I am willing to compute interannual variances  for the 
summer season July-August-September.
I first did that the most direct and simple way thanks to the 
transformations seasonal_reg@mod and seasonal_reg@modvar from the 
climatological_axes. In order to be sure of the result, I computed the 
variance manually, by first determining each of the 14 JAS seasons, the 
interannual mean of JAS, and then the variance, from the mathematical 
definition (and the N value).
Here is the problem that I have:
for one particular area that I am interested in, characterized by a 
interannual mean JAS of 7.6%, the first direct way gives me a variance 
of 12.7, over the 14 years, whereas the second and 'manual' method 
returns me a variance of 2,1...
I read in a former mail of the Ferret archive that the seasonal_reg 
transformation deals with missing values (I am using Ferret v5.51), and 
I have to precise that I have a month of september made of missing 
observations. Since I know what the manual method does, because it 
follows the definition of the variance, I am curious to know what 
exactly makes the transform seasonal_reg@modvar, and why there is such a 
big difference between the 2 ways of computing the variance...


Thanks in advance,
Helene


--
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Helene Brogniez
Post-Doctoral research associate / Climate Systems Center
Department of the Geophysical Sciences
The University of Chicago
5734 S. Ellis Ave.
Chicago, IL, 60637  USA

web: http://geosci.uchicago.edu/~hbrogniez