*** For details on how to be removed from this list visit the ***
*** CCP4 home page http://www.ccp4.ac.uk ***
Kay Diederichs wrote:
<snipped>
The sigmas of the datasets are not independant, rather they should be
similar (because the intensities should be the same, and strong
reflections have numerically high sigmaF's, so the sigma of a reflection
is correlated with its intensity).
Here is an example of the correlation between I and SigI for one of my
old data sets. As expected based on counting statistics, SigI is indeed
correlated with I.
>> give your option (or hit <return> to list options)
correl col 9 10
selected: CORREL
Resolution limits will be set to the limits
observed in the database
resolution limits used : 37.443676 - 1.63345063
Comparing columns :
IMEAN
SIGIMEAN
Grabbing statistics ...
$TABLE: Correlation vs. resolution:
$GRAPHS: Correlation vs. resolution:A:3,5:: R-factor vs. resolution:N:3,4:
$$
NSHELL DMIN DMAX RFACT CORREL <F1> <F2> NREF $$
$$
1 37.44 4.43 191.90 0.9297 1424.4 29.4 848
2 4.43 3.52 192.40 0.9521 1365.0 26.4 775
3 3.52 3.07 189.96 0.9292 929.3 23.9 763
4 3.07 2.79 189.38 0.9162 525.7 14.4 738
5 2.79 2.59 188.89 0.9289 386.2 11.1 735
6 2.59 2.44 188.82 0.9330 318.8 9.2 731
7 2.44 2.32 188.53 0.9261 280.1 8.3 733
8 2.32 2.22 188.38 0.9307 241.5 7.3 723
9 2.22 2.13 187.90 0.9126 209.7 6.6 714
10 2.13 2.06 187.40 0.8805 192.1 6.3 724
11 2.06 1.99 186.99 0.9074 158.0 5.3 709
12 1.99 1.94 185.80 0.8693 129.0 4.8 710
13 1.94 1.89 184.17 0.7946 101.4 4.3 715
14 1.89 1.84 182.05 0.7901 81.2 3.9 725
15 1.84 1.80 180.24 0.7362 65.4 3.5 689
16 1.80 1.76 176.87 0.6420 55.6 3.5 733
17 1.76 1.72 173.00 0.5187 41.0 3.2 688
18 1.72 1.69 171.37 0.5649 37.4 3.2 712
19 1.69 1.66 165.57 0.2884 29.9 3.1 705
20 1.66 1.63 157.98 0.2313 26.4 3.6 630
Based on counting statistics I would expect that the correlation would
be better between I and the variance of I (e.g. SigI**2). However, it
doesn't work out that way, which means I'm either missing something our
our sigma values contain considerable contributions beyond the counting
statitics contribution, and this additional contribution is more or less
proportional to I.
Here are the stats for I versus SigI**2
>> give your option (or hit <return> to list options)
correl col 9 16
selected: CORREL
Resolution limits will be set to the limits
observed in the database
resolution limits used : 37.443676 - 1.63345063
Comparing columns :
IMEAN
SigI2
Grabbing statistics ...
$TABLE: Correlation vs. resolution:
$GRAPHS: Correlation vs. resolution:A:3,5:: R-factor vs. resolution:N:3,4:
$$
NSHELL DMIN DMAX RFACT CORREL <F1> <F2> NREF $$
$$
1 37.44 4.43 116.38 0.5945 1424.4 3561.3 848
2 4.43 3.52 88.66 0.7352 1365.0 2018.8 775
3 3.52 3.07 103.94 0.6108 929.3 1921.0 763
4 3.07 2.79 79.20 0.5270 525.7 531.0 738
5 2.79 2.59 80.23 0.6956 386.2 302.4 735
6 2.59 2.44 82.18 0.7444 318.8 191.3 731
7 2.44 2.32 87.78 0.7102 280.1 147.5 733
8 2.32 2.22 92.07 0.7871 241.5 100.7 723
9 2.22 2.13 99.20 0.6261 209.7 93.5 714
10 2.13 2.06 102.97 0.5121 192.1 97.1 724
11 2.06 1.99 108.06 0.7500 158.0 50.5 709
12 1.99 1.94 123.41 0.5821 129.0 70.7 710
13 1.94 1.89 117.93 0.7167 101.4 28.4 715
14 1.89 1.84 116.50 0.7026 81.2 23.2 725
15 1.84 1.80 120.65 0.6890 65.4 18.2 689
16 1.80 1.76 114.66 0.6008 55.6 17.9 733
17 1.76 1.72 118.50 0.4996 41.0 13.6 688
18 1.72 1.69 114.22 0.4680 37.4 14.8 712
19 1.69 1.66 117.33 0.2521 29.9 12.9 705
20 1.66 1.63 118.10 0.1372 26.4 23.0 630
This should be done for a bunch of datasets but I wonder if someone else
has looked at this. Any ideas???
Bart
==============================================================================
Bart Hazes (Assistant Professor)
Dept. of Medical Microbiology & Immunology
University of Alberta
1-15 Medical Sciences Building
Edmonton, Alberta
Canada, T6G 2H7
phone: 1-780-492-0042
fax: 1-780-492-7521
==============================================================================
