Jesus, On Mon, May 23, 2016 at 2:07 PM, Jesús Eduardo Vega Castillo < jevecas@gmail.com> wrote:
Dear list,
I made an EXAFS analysis in Artemis (in Demeter 0.9.22) using all the available variables. This is a fragment of the .log file I got:
Independent points : 17.7666016 Number of variables : 17 Chi-square : 265.0947132 Reduced chi-square : 345.8050781
R-factor : 0.0026486
Number of data sets : 1 : k-range = 2.942 - 11.043 : dk = 1 : k-window = hanning : k-weight = 1,2,3 : R-range = 1.115 - 3.5 : dR = 0.0 : R-window = hanning : fitting space = r : background function = yes : phase correction = no : background removal = E0: 20002.215, Rbkg: 1.0, range: [2.25:17.4039986832903], clamps: 0/24, kw: 2 : epsilon_k by k-weight = 3.189e-004 : epsilon_r by k-weight = 2.339e-001 : R-factor by k-weight = 1 -> 0.00220, 2 -> 0.00224, 3 -> 0.00392
The problem is that when I use the Nyquist criterion
Nind=2*deltak*deltar/pi + 1
for calculating the number of independent points the value I got is much lower and close to 13.
I was not aware of this discrepancy and it caused a reviewer to think I made up the fit!
Sending the entire log file is always recommended. Otherwise, you are selectively editing what you are telling us, and expecting us to guess what you haven't told us. Sending the project file is often a good choice. Fortunately, you included an important clue. You have "background function = yes", which means the fit will actually extend down to R=0, using your Rmin as Rbkg in calculating how many variables are used for the spline. With that included, the fit does contain approximately 18 independent parameters, about 5 of which will be used for the background subtraction. --Matt