PDA

View Full Version : correlation



mahesh
25th August 2006, 09:10 AM
Hi ,
I need to know if we take the average windstree over some region and correlate with the precipitation or rainfall data which is not been averaged can we get the result?
I think we will get a point by taking the @ave of wind stress and the rain fall data is a field so help!!!

rocksea
25th August 2006, 09:40 AM
hi mahesh,

if both have same timeseries, we can do correlation analysis of the wind stress average at a region with rainfall at each points. so you will get correlation coefficients at each point over the rainfall region.

MoHDa
25th August 2006, 01:33 PM
This is a big story.
First of all, the description is not really enough to understand the problem that you want to solve, but I have the same impression with rocksea.
But, with this way you ll not get anything useful (i think so), it is better to start with cluster analysis (or any similar one) for both datasets that you want to correlate, to minimize the necessary amount of data (there are more benefits that this, but u ll realize them during the process).
A second thing, you have to find the mechanism (if exists) of the correlation:

Mathematical Correlation is a tool to give a number at a known Natural Correlation or to investigate the Correlation.

If you have no idea about the theory of the correlation, I suggest you to read any statistical book of social science, most of them are simplified but it is enough to understand the method(s), avoid the books of electrical engineering are good but to complicated.

Whatever you want, just ask...

mahesh
28th August 2006, 05:14 AM
hi mahesh,

if both have same timeseries, we can do correlation analysis of the wind stress average at a region with rainfall at each points. so you will get correlation coefficients at each point over the rainfall region.hi, thanx for replying
actually i've used the simple mathematical formulae to find correlation,
but in the end while plotting it is showing the value 0.


http://www.ferret.noaa.gov/Ferret/Mail_Archives/fu_2006/msg00403.html
u can check this link.

mahesh
31st August 2006, 10:10 AM
hello frends
I have posted one message abt correlation
" calculate the zonal wind stress at EQWIN[-ve zonal wind anomaly over equitorial region]
and correlate it with the rainfall"
so while finding the variance of wind stress it turned out to be 0, because variance of a single point is 0.
so the plot is showing the correlation of both is 0.
i got the temporal correlation but not the spatial correlation.
help

MoHDa
31st August 2006, 02:27 PM
Hello,

1. Do u use ur routines or any ready software?
2. Pls describe ur data, do u have for an area precipitation and wind? what's the spatial resolution? Do they have the same coordinate system?

Am sorry if u find my questions stupid but I don't know what u r trying to do.

For spatial correlation download from here (http://www.ai-geostats.org/geostats_faq/FAQ_software_suggestions.htm) one of the progs.
(http://www.ai-geostats.org/geostats_faq/FAQ_software_suggestions.htm)

yaswant
31st August 2006, 07:29 PM
hello frends
I have posted one message abt correlation
" calculate the zonal wind stress at EQWIN[-ve zonal wind anomaly over equitorial region]
and correlate it with the rainfall"
so while finding the variance of wind stress it turned out to be 0, because variance of a single point is 0.
so the plot is showing the correlation of both is 0.
i got the temporal correlation but not the spatial correlation.
helpI am sorry but I did not understand your problem completely! In computation of spatial correlation, the variance is calculetd over the spatail domain (not over a single point). In your case, the zonal wind anomaly and rainfall must be colocated and should have equal number of bins (samples; but this is not mandatory). Caution: the spatial relationship may not be significant, if your domain is small and both variables are spatially auto-correlated.

42d3e78f26a4b20d412==