Hi all,
If I have a set of variables for a set of samples [for example
absorbance readings from a spectrometer for a series of mixed
calibration solutions (partial factorial design) used to calibrate for
say Fe], and rank these variables according to their covariance, will
this ranking give me variables in order of maxumum variance and
greatest correlation with the analyte Fe? I am suspicious that using
this criterion has the following pitfall:
1. that if the standard deviation of the variable is large and the
correlation low, the high standard deviation will still give a high
covariance value due ot the high stdev - is this right?
Basically I need to be able to rank the variables on the following
criteria:
1. variables should have a high standard deviation,but we need
a measure not influenced by the size of the variable ( i.e we
have found that large intensity signals, due to large peaks, have
high standard deviations but are not directly correlated to the
analyte)
I thought that if maybe one could introduce into the covariance
equation the RSD term [stdev/mean], this would accomodate these
criteria, however I am not sure if this is possible, or if there is
another equation that will describe what I need.
Thanks for any help in advance
Mike Griffiths (3rd year Ph.D student: chemometrics and atomic
spectroscopy)
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|