[R-sig-Geo] newbie question in fitting data to surfaces

[Molins, Jordi]

Just an addition to the question I have asked before:

one easy way would be to fit:

- a global constant (to account for the average of the points of the
matrix);
- a plane to account for linear terms (in the example below, it should be a
plane with both negative partial derivatives with respect "x" and "y")
- a quadratic form in 2 dimensions to account for "curvature" terms
- maybe a cubic form

Is there an easy way to perform this, or some analogous solution?

Thank you

Jordi

------------------------------------


Hello,

I am new in this list. I have a problem and I think that the solution to
this problem can be found using spatial statistics, but I do not know how.

My problem is the following: I have a matrix of data, of around 6 rows and 5
columns. In reality, I have a time series, where at each point in time, I
have a matrix.

Usually (but not always) the maximum is at the position (1,1). Given any
position (i,j), the positions to the right and below are smaller numbers.
For example, 

21.75	21	20.3	18.9	17.4
21.1	20.5 ...
.
.
.
15.6	14.8	14.5	14.3	14.2

In my "world", if I have a row of data (1D, as opposed to 2D) I typically
fit a parameterized function to this data (e.g., theta_1 + (theta_2 +
theta_3 * x) exp(-x / theta_4), where x is the row of data). I get the
numbers theta_1, ..., theta_4 and I am happy.

However, now I have a matrix of data. Of course, I could fit a function for
each of the rows (or columns) and report the thetas for each row (or
column). But I guess that there should be a way to keep the dependencies
between rows (or columns).

Has anybody found a problem like this? Is there a predefined function (with
several thetas) that could cope with data like this one, and is easy to
calibrate? 

Thank you in advance

Jordi



--------------------------------------------------------------------------------
The information contained herein is confidential and is inte...{{dropped}}