datcmp bug?

[jbhopkins]

Hi folks,

I was playing around with datcmp and think I may have found a bug. I took a CRYSOL fit, pulled out the data and fit, and ran datcmp with the chi^2 test:

(base) jessehopkins@medusa chi_sq % datcmp --test=CHI-SQUARE 2POL_polymerase_FIT.dat 2POL_polymerase.dat
Hypothesis: all data sets are similar
Alternative: at least one data set is different

Pair-wise Reduced Chi^2 test with correction for Familywise Error Rate (Bonferroni)
Chi^2 Pr(>Chi^2)adj Pr(>Chi^2
1 vs. 2 0.760088 0.999907 0.999907

1* 2POL_polymerase_FIT.dat
2* 2POL_polymerase.dat


The CRYSOL .fit file has the chi^2 as 1.52, and I wrote a quick little python calculator for reduced chi^2 and got 1.52 as well. So there seems to be a factor of 2 off somewhere in the chi^2 calculator?

Data are attached.

Jesse

[jbhopkins]

I think I figured out what's going on. The extracted fit data had the experimental uncertainties associated with it (to make a 3 column dataset, for lack of anything else to use). If instead of the experimental uncertainties being used to calculate chi^2 I use uncertainties that are created by summing the uncertainties for two datasets in quadrature I get the result that datcmp is outputting.

So it seems like for two datasets with specified uncertainties datcmp does a quadrature sum and uses that as the uncertainty for chi^2. Is there a statistical reason for that? I've never encountered that before.