PCA for NIR Spectra_part 006: "Mahalanobis"

February 28, 2012
By

(This article was first published on NIR-Quimiometría, and kindly contributed to R-bloggers)

Outliers have an important influence over the PCs, for this reason they must be detected and examinee.
We have just the spectra without lab data, and we have to check if any of the sample spectra is an outlier ( a noisy spectrum, a sample which belongs to other population,……., an extreme sample for a particular constituent,..).
One way to detect outlier is the “Mahalanobis distance”. And we are going to calculate the Mahalanobis distance to the center of the population.
Previously we have calculated with the NIPALS algorithm the T score matrix, for three PCs.
Let´s calculate now the Mahalanobis distance (Library: Chemometrics) for a certain probability.

> Moutlier(T3pc_nipals,quantile =0.975, plot=TRUE)
$md [1] 3.4506943 2.0677806 1.8554593 1.3493939 1.0093176 1.4537839 1.7177865 [8] 0.8551981 1.9936152 1.7136093 1.6755040 1.0370122 0.6980203 1.0457686 [15] 1.9107813 2.4836284 1.8483303 1.6226210 1.7422189 2.3278939 2.9407373 [22] 1.0270963 0.8200505 1.3194672 1.0637166 0.7383098 0.9658263 1.1470021$cutoff
[1] 3.057516

We can see how the sample number “1” is out of the cutoff ratio.

To leave a comment for the author, please follow the link and comment on his blog: NIR-Quimiometría.

R-bloggers.com offers daily e-mail updates about R news and tutorials on topics such as: visualization (ggplot2, Boxplots, maps, animation), programming (RStudio, Sweave, LaTeX, SQL, Eclipse, git, hadoop, Web Scraping) statistics (regression, PCA, time series, trading) and more...

If you got this far, why not subscribe for updates from the site? Choose your flavor: e-mail, twitter, RSS, or facebook...

Tags:

Comments are closed.