Description Usage Arguments Details Value Author(s) See Also Examples
By definition the total ion count of a spectrum is the sum of all the m/z signal intensities (sum over the columns). A robust regression can be built to model the effect of the injection order on the TIC of each sample. The fitting residuals are used to evaluate the median of the absolute deviations (MAD) from the linear equation. Sample TIC and linear model are then plotted according to the injection order to identify potential outliers and or structure in the data.
1 |
x |
A numeric data frame or matrix to be processed. |
injorder |
A numeric vector corresponding to the injection order of each sample. |
thres |
A numeric value of threshold for detecting outlier. |
As an easy diagnostic measure, the TIC can provide an estimation of factors
that may affect the overall intensity of the run such as gradual instrument
drift (e.g. resulting from loss of sensitivity of the ion source), or step
changes in instrument characteristics after maintenance. Also, an examination
of the TIC can reveal suspicious samples where unusually low or especially
high signal intensities in some runs may be due to contamination or poorly
extracted samples. A regression can be built to model the effect of the
injection order on the TIC of each sample. As a conservative rule, any sample
that deviates more than 2/3 (argument thres
) times from the MAD must be
examined manually to identify the origin of the different intensity behaviour
and then removed before further statistical analysis if corrective measures do
not improve the individual fingerprint. Further assessment of outlying samples
is discussed later. In any case where a linear relationship (i.e. gradually
changing TIC in sample set) is observed between the injection order and sample
TIC, this dependency will be removed by TIC normalisation. If other structure
related to the order of injection is noticed, for example an analytical batch
effect (i.e. a step change in TIC at beginning or in middle of an injection
series), the user must identify its potential origin (e.g. changes in machine
calibration or mobile phase) and possibly create a new experimental factor
(batch) where each step change in level corresponds to the start of a new batch.
A list containing the following components:
resid |
Sample residuals. |
mod |
Robust linear model. |
loutl |
List of outlying samples as defined by the
|
David Enot and Wanchang Lin dle,wll@aber.ac.uk
rlm
.
1 2 3 |
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.