The sign of the numbers indicates correlation vs. anticorrelation. For example, Uiso's correlate between atom 1 and 2 in phase "1 1 1UISO" & "1 2UISO" so that Uiso increasing on one atom has a similar impact as an increase on the other. In the case of the reciprocal cell tensor (related to cell parameters) an increase on one of the diagonal terms has a similar effect as a decrease on the zero correction ("1 RM11" & "ZERO 1").

Brian


On Apr 8, 2008, at 8:55 AM, Leopoldo Suescun wrote:

Dear Gerard,
Here is how to read it:

The correlation matrix terms greater than 0.50 are
1  1UISO-1  2UISO   5011  3UISO   687   1HSCL   801
1  2UISO-   1HSCL   650
1  3UISO-   1HSCL   852
1   RM11-ZERO  1   -642
1   RM22-ZERO  1   -640
1   RM33-ZERO  1   -692
   1ABS1-   1ABS2  1000
1  1PF 2-1  1PF 3  -727
v-cov distribution 0...1 by 0.05
   430  125   77   34    7   10    7    2    0    1    1    0    2
  3    1    0    1    1    0    1

1st column (8 characters wide): Variable 1, 2nd column: (also 8
characters wide) Variable 2, with "-" separating both columns, 3rd
column (6 characters wide) the correlation value multiplied by 1000.
Following collumns are show if more paramters have correlations greater than 0.5 with the parameters in the first collumn. So in this case, 4th
collumn (8 characters wide with no separation from 3rd): Variable 2',
5th collumn (6 char wide): correlation between Var1 and Var2', etc for
var 2''...

So you can read the first column this way:
1 1UISO (Iso thermal param. for atom 1 in phase 1) shows a correlation
of 0.501 with 1  2UISO (iso param for atom 2 in phase 1) so thermal
parameters for atoms 1 and 2 inphase one have 50% correlation.
1  1UISO shows .687 corelation with 1   3UISO, or atom's 1 and 3 Uisos
are correlated 68.7%
1  1UISO is correlated with histogram scale factor by 80.1 %

ABS = absorption correction if my memory is good but you can check in
the GSAS manual that accompanies GSAS distribution. You have a problem
with the setup of the refinement because you have 2 absorption
corrections and both parameters are correlated by 100% so they are not
refining independently. I don't remember what the numbers before and
after ABS mean (histogram, phase, model applied) but check in the manual
and you'll find it.

The last line, v-cov distribution 0...1 by 0.05 shows the values in a
histogram with range of covariances from 0 to 1 divided in classes 0.05 wide. First value means that 430 pairs of parameters show absolute value
of correlation between 0.0 and 0.025, 2nd value means 127 parameters
show absolute value of correlation between 0.025 and 0.075 and so forth.

The top list only gives those higher than 0.5 in absolute value so you
expect 10 counting the last 10 entries in the bottom line, as shown.

Hope this helps.
Leo

Dr. Leopoldo Suescun
Argonne National Laboratory
Materials Science Division Bldg 223
9700 S. Cass Ave. Argonne IL 60439
Phone: 1 (630) 252-9760
Fax: 1 (630) 252-7777

----- Original Message -----
From: "Gerard, Garcia S" <[EMAIL PROTECTED]>
Date: Tuesday, April 8, 2008 7:37 am
Subject: correlation matrix interpretation

Dear all,

Could you please help me to translate this chunk of the correlation
matrix?HSCL is the histogram scale factor
PF profile coefficients
RMii lattice parameters
ABS?
UISo thermal parameters

how do i interpret the numbers? and the signs? how strongly
correlated these variables are?

regards




--
Heriot-Watt University is a Scottish charity
registered under charity number SC000278.


<leopoldo.vcf>

********************************************************************
Brian H. Toby, Ph.D.                            office: 630-252-5488
Materials Characterization Group Leader, Advanced Photon Source
9700 S. Cass Ave, Bldg. 433/D003             work cell: 630-327-8426
Argonne National Laboratory         secretary (Marija): 630-252-5453
Argonne, IL 60439-4856         e-mail: brian dot toby at anl dot gov
********************************************************************


Reply via email to