Questionable multivariate statistical inference in wildlife habitat and community studies

Eric A. Rexstad; Dirk D. Miller; Curtis H. Flather; Eric Anderson; Jerry W. Hupp; David R. Anderson

Treesearch

Questionable multivariate statistical inference in wildlife habitat and community studies

Formally Refereed

Authors:

Eric A. Rexstad, Dirk D. Miller, Curtis H. Flather, Eric Anderson, Jerry W. Hupp, David R. Anderson

Year:

1988

Type:

Scientific Journal

Station:

Rocky Mountain Research Station

Source:

The Journal of Wildlife Management. 52(4): 794-798.

Download (PDF 848 KB)

Download (PDF 848 KB): https://research.fs.usda.gov/download/treesearch/59195.pdf

Abstract

We analyzed a data set constructed from functionally unrelated, easily collected observations (e.g., meat, stock, and liquor prices) around Fort Collins, Colorado, using principal components analysis (PCA), canonical correlation analysis (CC), and discriminant function analysis (DFA). Each produced seemingly significant results and suggested strong relationships between the variables measured. We suggest that multivariate techniques can provide invalid inferences when used with data containing no relationships. We question the use of these techniques in studies of wildlife habitat.

Keywords

wildlife habitat, data sets, principal components analysis (PCA), canonical correlation analysis (CC), discriminant function analysis (DFA), multivariate techniques

Citation

Rexstad, Eric A.; Miller, Dirk D.; Flather, Curtis H.; Anderson, Eric M; Hupp, Jerry W.; Anderson, David R. 1988. Questionable multivariate statistical inference in wildlife habitat and community studies. The Journal of Wildlife Management. 52(4): 794-798.