| Autor: |
David H. Burns, Francis W. L. Esmonde-White |
| Rok vydání: |
2004 |
| Předmět: |
|
| Zdroj: |
Vibrational Spectroscopy. 36:287-292 |
| ISSN: |
0924-2031 |
| DOI: |
10.1016/j.vibspec.2004.02.012 |
| Popis: |
A method for applying two-dimensional correlation analyses to N-way data is presented, which compactly displays trends in the data. Interpreting large multi-dimensional data sets obtained from hyphenated instruments is difficult. Several methods of applying two-dimensional correlation analyses to N-way data are examined. To visualize trends occurring along the original parameters, data are stacked lexicographically according to N−1 parameters in preparation for synchronous and asynchronous correlation analyses (2D COSY). Correlation analysis comparing a first physical parameter to all other parameters indicates trends. The average correlation intensity indicates significance of the correlation. It does not indicate variation between axes. Standard deviation of correlation intensity is used to determine the magnitude of variation about the mean. This shows the extent correlation changes with other parameters. Pearson correlation coefficients are computed. The sign of correlation coefficients indicates trend direction. Combining the sign of Pearson correlation coefficients with standard deviation magnitudes indicates magnitude and direction of correlation changes. These visualization methods are used to investigate mathematically generated three-way data. Results show that a small number of figures can illustrate variation in the signal and isolate the components causing signal variation. This provides a rapid means for initial trend-screening of large multi-dimensional data sets. Rapid trend screening is important for both data acquisition and data analysis. |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect