Hi, everybody...
I do my research with using FTIR analysis. I measure normal and cancer cells. All measures I am doing in reflectance mode with using the low-e glass as the substrate for cells.
I read a lot of articles about using different kinds of statistic for normalization and calculation of data after measurements.
I would like to know, what is the right succession of data processing after measurements.
In general, I use RMieS-EMSC correction algorithm and after I calculate the average spectrum and then I calculate the height and area of the main peaks in this spectrum. In the articles, I read that after using RMieS-EMSC correction algorithm, some authors use PCA-LDA analysis but the do it with 2th derivative of spectrums. What is the difference of using PCA-LDA analysis after RMieS-EMSC correction or after 2th derivative? What statistic program is more easiest to use for PCA-LDA analysis? How to explain the results PCA-LDA analysis? Maybe, somebody has a large experience in processing of FTIR research of cells...I will be appreciated if somebody will give me the information how is correct to do such calculations step by step...
Thank you...
Dear Serdiuk
Data from FTIR or other spectral instruments analysed by different chemometric methods. There are three several steps of this analyses, first step is pre-processing of spectral data using correction methods or derivative functions. There are lots of method for signal and baseline correction in spectral data. Selection of true pre-processing method is related to problems in spectral data. You should consider the problems (signal or baseline) in your data set when you select the pre-processing method. Second step of the analysis, wavelength selection and it enable to user using "true variables" in the model development. Actually, variable selection is not mandatory, but it had a significant effect on prediction power of the generated models. The last step is model development and validation. PCA-LDA analysis is combined of principal component analysis with linear discriminant analysis. This analysis can be perdormed in Unscrambler (CAMO, Norway), which is one of the best software for spectral data analysis. There are several Matlab toolbox and you can use these toolboxes in your analysis. But I suggest the Unscrambler, because of it is a user friendly and easy to use software for such data.
Sincerely
Fatih Kahrıman
Indeed, the preprocessing of the spectral data is a very complex issue in particular since the quantity that you measure (reflectance) is not directly proportional to the number or concentration of the molecular species in your sample. Spectrometer give the illusion it would be the case by using a conversion to absorbance that can be highly erroneous (like A = -lgT or A = -lg R), possibly spoiling any further effort to obtain a meaningful statistical analysis. Unfortunately, this has not been investigated in literature so far.
Hi, I'm also interested in knowing how to use the LDA-PCA analysis. I'm using Unscrambler software. I will be appreciated if somebody will give me the information.
Thank you
One addition: If you measure in reflectance and use low-e glass, i.e. in transflection, your spectra are susceptible to the "electric field standing wave effect". Before any further spectral processing makes sense, you must correct your spectra for this effect, which RMieS-EMSC is not able to do! See Article The electric field standing wave effect in infrared transfle...
Article Removing interference-based effects from the infrared transf...
Hello Vitalii Serdiuk . As Fatih Kahrıman suggested you, Unscrambler (CAMO) is a very convenient software and especially when you would like to have quick idea of your spectra. However, when you are preforming a LDA, I can suggest you to use the SPSS IBM. With this software, you can go farther in the understanding and the interpretation of the statistical results.
- Badges
- Science topic
- Similar topics
- Correction