Kolmogorov–Smirnov Test for Statistical Characterization of Photopyroelectric Signals Obtained from Maize Seeds
- Authors
- Type
- Published Article
- Journal
- International Journal of Thermophysics
- Publisher
- Springer US
- Publication Date
- Nov 07, 2018
- Volume
- 40
- Issue
- 1
- Identifiers
- DOI: 10.1007/s10765-018-2462-4
- Source
- Springer Nature
- Keywords
- License
- Yellow
Abstract
Photothermal techniques are useful experimental methodologies for characterization of the optical and thermal parameters of different materials like maize seeds due to its advantages such as non-invasive and non-destructive nature. Among these techniques, the photopyroelectric microscopy was applied in the present research to obtain thermal images where each of their coordinates represents amplitude values of the photopyroelectric signal, indicating differences in the structural components of both genotypes of maize seeds. The random variations of the amplitude of the photopyroelectric signal caused by the heterogeneous nature of the thermal properties of the samples, were represented by histograms to identify the probability density function underlying the data sample, observing that in the case of the maize seed with floury structure, the amplitude variations could be described statistically by the transformed Moyal distribution when a linear transformation with censored data was applied to the data set obtained from the thermal image with a significance level of 0.001, according to the Kolmogorov–Smirnov statistical test for goodness-of-fit. In the case of the photopyroelectric signal obtained from a maize seed with crystalline structure, it was not possible to describe statistically the amplitude variations of the signal by means of the transformed Moyal distribution because it did not pass the Kolmogorov–Smirnov test, so the same statistical test for goodness of fit was applied to both genotypes of maize seeds for analyzing the populations of the data sample, in order to find the distributions that best fit each population with a significance level of at least 0.05 increasing in this way the power of the test. The distributions with the best fit were logistic, log-logistic, uniform, least extreme value and normal.
