A discussion on the significance associated with Pearson’s correlation in precision agriculture studies

• Published in: Precision agriculture 2013-10, Vol.14 (5), p.558-564
• Main Authors: Taylor, J. A., Bates, T. R.
• Format: Article
• Language: English
• Subjects: Agricultural research; Agriculture; Atmospheric Sciences; Biomedical and Life Sciences; Canopies; Chemistry and Earth Sciences; Computer Science; Correlation analysis; Life Sciences; Physics; Precision farming; Remote Sensing/Photogrammetry; Sensors; Short Discusssion; Soil Science & Conservation; Spatial data; Statistics for Engineering; Wineries & vineyards
• ISSN: 1385-2256; 1573-1618
• DOI: 10.1007/s11119-013-9314-9

Pearson’s correlation is a commonly used descriptive statistic in many published precision agriculture studies, not only in the Precision Agriculture Journal, but also in other journals that publish in this domain. Very few of these articles take into consideration auto-correlation in data when performing correlation analysis, despite a statistical solution being available. A brief discussion on the need to consider auto-correlation and the effective sample size when using Pearson’s correlation in precision agriculture research is presented. The discussion is supported by an example using spatial data on vine size and canopy vigour in a juice-grape vineyard. The example data demonstrated that the p -value of the correlation between vine size and canopy vigour increased when auto-correlation was accounted for, potentially to a non-significant level depending on the desired α-level. The example data also demonstrated that the method by which data are processed (interpolated) to achieve co-located data will also affect the amount of auto-correlation and the effective sample size. The results showed that for the same variables, with different approaches to data co-location, a lower r -value may have a lower p -value and potentially hold more statistical significance.