Frequently Asked Questions (FAQ)
Can the mat2corr toolbox account for unevenly spaced sampling of spectra?
Unfortunately, the current version of the mat2dcorr toolbox (v. 1.05) does not consider non-equidistant vectors of the perturbing variable. However, this would be a good idea for future versions of the toolbox as an addition to its functionality.
In the literature, there are basically three different ways to deal with perturbations that do not fulfill the equidistance condition:
- 1. Ignore the requirement for equidistant perturbation vectors and use the data as they are. This is what the current mat2dcorr toolbox version 1.05 does.
- 2. Use modified correlation equations as described in Noda (2003) to account for uneven sampling of the perturbation variable.
- 3. Interpolate the perturbation values and the associated spectral data to get an equidistant distribution of the perturbation values.
Option 3 would be fairly easy to implement programmatically. It is already on the to-do list.
I have loaded spectral data, but the buttons are still grayed out?
This is not a bug! It is important to understand that the mat2dcos toolbox was originally developed for heterospectral 2D correlation analysis (2D-COS). This means that spectral series from two different modalities are analyzed. For example, if IR and Raman data are to be analyzed by heterospectral 2D-COS, two different spectral series must be loaded into the toolbox. For the probably more common case of autocorrelation 2D-COS, this means that the data set to be analyzed has to be loaded twice, once as an x- and a second time as a y-data set. Only then will the buttons be available for analysis.
How to interpret intensity values of 2D correlation functions?
For example, the symmetric 2D function can be interpreted as a statistical value, namely the covariance between two variables. Unlike the correlation, the covariance is not normalized by the standard variation and can thus take very small or very large values. This depends, among other things, on the absolute values of the intensity changes in the spectra examined. For this reason, some researchers do not attach much importance to the absolute intensity values of the 2D-COS functions. It is more important that the maxima and minima of the 2D correlation functions are clearly above the noise, which can be checked by analyzing the correlation slices.
It is also noteworthy that the intensities of the 2D functions differ depending on the method used to calculate the 2D correlation functions (e.g. statistical or FFT based). This can be easily tested using the example data provided with the mat2dcorr toolbox. The different intensities obtained by the distinct 2D-COS functions are not due to programming errors and are not discussed in the scientific literature.
Links to other non-commercial 2D-COS software solutions
- 2DShige, free software for two-dimensional correlation analysis developed by Prof. Shigeaki Morita (Osaka Electro-Communication University).
- MIDAS 2010, Matlab-based software tools developed in the Canadian Light Source for 2D spectroscopic analysis and data exploration of time resolved infrared spectra
- corr2D (R), - Implementation of two-dimensional correlation analysis in R, developed by Robert Geitner.