So, if I have multiple FTIR spectra each taken on different location of the sample and averaged from ~20 spectra, how can I check if the sample is homogeneous, i.e., if there are statistically significant differences between the spectra based on location? For visuals, I tried to compute at each wavenumber confidence upper and lower limits, but I am not sure if this method is valid even for visualisation and it does not give anything exact or quantitative.

  • $\begingroup$ It is a statistical problem, which can be solved with maybe 95% degree of confidence, using the chi-squared test. $\endgroup$ – Maurice Sep 2 '20 at 11:51
  • $\begingroup$ Without seeing the spectra it is hard to advise , but presumably you have a set of peaks that you expect to find. If you have a 'standard spectrum' then use this to compare peaks of others and so remove poor samples. Plotting this spectrum on x vs each of the rest on y and making a contour plot may help to do this pairwise. As the spectra are likely to be so different due to extraneous (unimportant) lines a conventional stats method may not work. $\endgroup$ – porphyrin Sep 2 '20 at 12:17
  • $\begingroup$ @Maurice Here I tried to construct chi squared test in R, but got a very large value: stackoverflow.com/questions/63714065/… . $\endgroup$ – user Sep 2 '20 at 22:01
  • $\begingroup$ @ User. Well ! What can I say ? I don't know your numerical values. And even if I would have known them, I could not change the chi squared value... $\endgroup$ – Maurice Sep 3 '20 at 13:49
  • $\begingroup$ @Maurice Ok, the problem was that I haven't divided everything by N. $\endgroup$ – user Sep 3 '20 at 14:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.