A Standardized Analysis System

This week has been another week of laser analysis, and it looks like I will finish all of the laser work by next Friday. I'm excited, since it means I'm coming to the true end of this section of my research, which has spanned a year-and-a-half now. The last few steps I have to complete in this project are analyzing the data and collecting some final measurements. A few weeks ago I mentioned how I upload the data points from the laser analysis and compare those points to known samples of plastics and other items to determine what type of material I have scanned. The two pieces of information I learn from this process are the polymer (or other material) type, and the correlation coefficient. However, these two data points do not amount to the end of the polymer analysis step. I need to, at last, choose an appropriate cutoff value for the correlation coefficient to determine whether I include a given particle in my data analysis.

Since the Raman laser that I am using is designed for larger particles, there is a higher likelihood for error in my research, which is why it is so important that I use cutoff values that are supported by microplastics literature. Therefore, although the spectral analysis software may identify a microparticle as plastic, I need to compare the particle's coefficient to my cutoff value to determine whether or not I include the particle in the analysis phase. The literature, by the way, suggests that a coefficient of 0.7 (out of 1.0) is an appropriate cut-off value and a scientist can visually compare particles with coefficients between 0.6 and 0.7 to determine their inclusion in the work. I have been worried about this step since the first time I used the Raman machine, because I knew that the laser I have access to is not meant for such fine scale processing and therefore the spectra may not be 100% representative of their chemical composition. Another source of unreliable spectral analysis is the use of colorants or additives in the plastic production process. Some dyes will reflect the laser more apparently than the plastic itself, and therefore the analysis software may register the particle as a colorant, rather than a plastic. 

There are two ways that I can assess the reliability of the spectral analysis to ensure that the data are not biased by the analysis method. The first is the search function in OpenSpecy's software, which allows me to search for a potential polymer comparison to the unknown microparticle. OpenSpecy will bring up the spectra for the known polymer and provide the associated correlation coefficient, which helps me assess whether I need to re-run the laser analysis. If the correlation is exceptionally low, there's a strong chance that the material is not plastic at all. Notably, I have completed this step a few times and found that the correlation is often non-existent, which suggests that the current analysis is appropriate. The second way I can assess reliability is through the use of statistics to check if certain colored particles are inexplicably removed due to their pigmentation. If the coloration of a particle is frequently showing the same non-plastic material or colorant, then perhaps there is a bias in the analysis method. 

Since I don't have all the data yet, I cannot speak on either of these two issues, though they will be concerns that I address when I finish writing the manuscript associated with this research. In the hopes of providing new evidence in a rather young field, I do not want to rush the analysis phase or make mistakes that I can avoid by slowing down and running some 'due diligence' steps.

Next week will potentially be the last week for a new post about this research. Come with me to see what happens when the final particle is processed and I have crossed all the t's and dotted the i's. Note: the blog will not end because this phase of research ends. There's a lot more work that I will be doing and I can't wait to bring you along.