Choice of variables for correlation analysis when measurement error is present

F. Thomas Lloyd

Treesearch

Choice of variables for correlation analysis when measurement error is present

Informally Refereed

Authors:

F. Thomas Lloyd

Year:

1978

Type:

Scientific Journal

Station:

Southern Research Station

Source:

Proceedings Soil Moisture-Site Productivity Symposium USDA Forest Service Southeastern Area S&PF

Download (PDF 181 KB)

Download (PDF 181 KB): https://research.fs.usda.gov/download/treesearch/59804.pdf

Abstract

Although correlation analysis is widely used, it is not well known that using variables measured with error causes the correlation coefficient to be underestimated (or biased toward zero). Knowledge of the form of this bias will help in the selection of variables to be used in studying soil and site productivity relationships . An example using site index shows that bias can be appreciable.

Keywords

Bias, site index, variance ratios, Pinus taeda

Citation

Lloyd, F. Thomas. 1978. Choice of variables for correlation analysis when measurement error is present. In: Proceedings Soil Moisture-Site Productivity Symposium USDA Forest Service Southeastern Area S&PF; 1978; Atlanata, GA: 292-295.