The effects of measurement errors, collinearity and their interactions on aerosol source apportionment computations

Meng dawn Cheng, Philip K. Hopke, Dennis E. Jennings

Research output: Contribution to journalArticlepeer-review

6 Scopus citations

Abstract

Cheng, M.D. Hopke, P.K. and Jennings, D.E., 1988. The effects of measurement errors, collinearity and their interactions on aerosol source apportionment computations. Chemometrics and Intelligent Laboratory Systems, 4: 239-250. The effects of analytical precision of chemical analyses of aerosol sources and ambient samples and of collinearity between sources on the solution of a generalized chemical mass balance (CMB) model for aerosol mass apportionment were examined. The analytical imprecision (in the range of 5-10 percent in terms of noise-to-signal ratio) was found to be a key factor in determining the precision and accuracy of the CMB source apportionment if and only if there is moderate or no source collinearity among the source composition profiles. This result implies that the CMB solution may be unreliable at higher noise-to-signal ratios (i.e. larger than 10 percent) under even the mild collinearity conditions used in this study. At the other end of this simulation, under a severe collinearity, the outcome of the CMB aerosol apportionment calculation was unpredictable because the extreme collinearity dominated the other controlling factors. The use of the effective variance weighted (EVW) least squares scheme did not yield more precise and accurate estimates for source contributions than the ordinary weighted scheme in this study even when the underlying assumption of normality (Watson et al., Atmospheric Environment, 1984) for the EVW scheme was true.

Original languageEnglish
Pages (from-to)239-250
Number of pages12
JournalChemometrics and Intelligent Laboratory Systems
Volume4
Issue number3
DOIs
StatePublished - Sep 1988
Externally publishedYes

Fingerprint

Dive into the research topics of 'The effects of measurement errors, collinearity and their interactions on aerosol source apportionment computations'. Together they form a unique fingerprint.

Cite this