The Probability of Chance Correlation Using Partial Least Squares (PLS)

Matthew Clark, Richard D. Cramer

Research output: Contribution to journalArticlepeer-review

287 Scopus citations

Abstract

The frequency of chance correlation using partial least squares (PLS) has been measured experimentally for variously dimensioned data, comprising either completely random numbers, random numbers containing a perfect correlation within, and CoMFA field descriptors. This frequency, much lower than that for stepwise multiple regression, is maximal for datasets in which the number of descriptors equals the number of compounds, and surprisingly decreases indefinitely as the number of descriptors becomes much greater than the number of compounds. However, perfect correlations involving descriptor subsets are not detected by PLS if the number of irrelevant descriptors is excessive. In CoMFA applications, the probability of chance correlation is usually negligible. For example with 21 compounds a crossvalidated r2 value greater than 0.25 will occur by chance in less than 5% of trials.

Original languageEnglish
Pages (from-to)137-145
Number of pages9
JournalQuantitative Structure‐Activity Relationships
Volume12
Issue number2
DOIs
StatePublished - Jan 1 1993

Keywords

  • chance correlation
  • CoMFA
  • cross validation
  • Partial least squares
  • stepwise regression

Fingerprint

Dive into the research topics of 'The Probability of Chance Correlation Using Partial Least Squares (PLS)'. Together they form a unique fingerprint.

Cite this