Chi^2 missing in CRYSOL output

Calculation of SAXS and SANS profiles (CRYSOL, CRYSON), superposition of models (SUPCOMB, DAMAVER, DAMCLUST), database DARA
Post Reply
Message
Author
aleful
Member
Posts: 6
Joined: 2016.11.18 18:26

Chi^2 missing in CRYSOL output

#1 Post by aleful » 2016.11.18 18:39

In order to compare theoretical scattering from xtal structure with experimental data from solution I used CRYSOL. Although I received .fit output file with two curves, the chi^2 value is not reported. What is the way to have it printed? The log file in attachment.

thank you in advance,
Attachments
dimer_model00.log
(3 KiB) Downloaded 69 times

User avatar
AL
Administrator
Posts: 651
Joined: 2007.08.03 18:55
Location: EMBL Hamburg, Germany
Contact:

chi-square is too large

#2 Post by AL » 2016.11.18 19:20

You may check 'crysol_summary.txt' and the header of the .fit file but basically your chi-square is way too large to make sense, this is why you see '*******' instead of a number.
You can recalculate chi-square of the fit using datcmp:

Code: Select all

$ datcmp --test=CHI-SQUARE dimer_model00.fit

aleful
Member
Posts: 6
Joined: 2016.11.18 18:26

Re: Chi^2 missing in CRYSOL output

#3 Post by aleful » 2016.11.22 13:49

Thank you for the answer. Do you know what is the range of Chi^2 "too high to be displayed"?. When I use datcmp the calculated Chi^2 is -1, what is a ridiculous value.

I am aware that the fit is far from perfect, because xtal structure is lacking 52 residues at N-terminus (out of 133 contributing to scattering in solution); still I would like to have the numerical value to show the improvement after including the missing part in the model.
Attachments
dimer_model00.fit
(36.77 KiB) Downloaded 59 times

franke
Administrator
Posts: 407
Joined: 2007.08.10 11:09
Contact:

Re: Chi^2 missing in CRYSOL output

#4 Post by franke » 2016.11.22 22:15

aleful wrote:Thank you for the answer. Do you know what is the range of Chi^2 "too high to be displayed"?. When I use datcmp the calculated Chi^2 is -1, what is a ridiculous value.
Not ridiculous. Lemme guess, no error estimates? Then that's what you get. Get accurate errors, and datcmp gives you an accurate chi^2.

With respect to the range: depending on the number of points of experimental data, the chi^2 value should be in the rule-of-thumb range of 0.9 to 1.1 (assuming ~1500 experimental points, a slightly wider range if less, more narrow if more). Of course, this assumes correct errors. Without correct errors, all other bets are off, too. And with accurate errors "*****" would indicate, roughly translated, "too embarrassingly high to even show the exact numerical value as it has no further meaning here", without errors it may mean anything. Probably division-by-zero. Or Not-A-Number. Or maybe the winning numbers of next weeks lottery.

aleful
Member
Posts: 6
Joined: 2016.11.18 18:26

Re: Chi^2 missing in CRYSOL output

#5 Post by aleful » 2016.11.23 19:10

Thank you for your answer. By saying ridiculous I meant I expect any value > 0. Especially that I have error estimates in my .dat file (column entitled sig(q), if I understand correctly). However, I cannot evaluate how accurate they are, but they are always at least one order of magnitude smaller than the corresponding intensity. The .dat file in the attachmnent.
Attachments
dimer_edited.dat
(194.97 KiB) Downloaded 53 times

ckerr
Active member
Posts: 86
Joined: 2015.03.25 09:03
Location: EMBL Hamburg

Re: Chi^2 missing in CRYSOL output

#6 Post by ckerr » 2016.11.24 10:01

What version of ATSAS are you using? A recent version should copy the sig(q) column into the .fit file so that it can be used to calculate the chi^2 - but your .fit file has no sigma column.
Last edited by ckerr on 2016.11.24 10:02, edited 1 time in total.
Reason: ATSAS version not CRYSOL version

User avatar
AL
Administrator
Posts: 651
Joined: 2007.08.03 18:55
Location: EMBL Hamburg, Germany
Contact:

ATSAS version?

#7 Post by AL » 2016.11.24 10:19

Which ATSAS version do you have? Please check what you get if you execute

Code: Select all

crysol --version
Starting from ATSAS version 2.7.0 CRYSOL does not skip the experimental errors from the fit file by default.

aleful
Member
Posts: 6
Joined: 2016.11.18 18:26

Re: Chi^2 missing in CRYSOL output

#8 Post by aleful » 2016.11.24 16:37

crysol, ATSAS 2.7.1 (r6669)

it was downloaded and installed not more than a month ago

ckerr
Active member
Posts: 86
Joined: 2015.03.25 09:03
Location: EMBL Hamburg

Re: Chi^2 missing in CRYSOL output

#9 Post by ckerr » 2016.11.24 17:52

The latest version is 2.7.2 so I suggest that you upgrade. Then try running crysol and datcmp again and tell us if the problem still appears.

UPD: ATSAS 2.8.0 is out now.

Post Reply