Back to the Top
I would be very thankful if anyone could offer a step by step algoritm
for application of weightin procedure for calibration curves. It would
be very helpful, if one could define when wgeighting is legitimite and
what type of wheighting coifficient need to be applied.
Thank you.
Sincerely, Tahir Kasumov, PhD
CCF
Back to the Top
Dear Tahir Kasumov,
As generally calibration curves are plotted using =20
Least squares regression, Weighting minimizes the sum of squares of =20
the relative distance of data from the curve. Generally 4 kinds of =20
weightings will be used for Calibration curve (1/x, 1/x2, 1/y, and 1/=20
y2). Selection of weighting factor solely depends upon the number of =20
points that you select at LOQ/ULQ end. If the variation in dependent =20
variable is more towards ULOQ end then you have to go for 1/y or 1/y2. =20=
1/x or 1/x2 weighting will be selected when we want to weigh the =20
points more towards left point of the graph i.e., at LOQ end.
The formulas for calculating the weighting are
1/x -> 1/xdata (xdata-xcurve)2
1/x2 -> 1/xdata2 (xdata-xcurve)2
Similarly the same formulas are applicable to =93Y=94 weighting also. =
More =20
often =93y=94 weighting will be used and =93X=94 weighting will be used =
rarely =20
as there will be more necessity to weigh the points at the ULQ end due =20=
to broader ranges (less no .of points at ULQ end).
With Regards,
Veeravalli Vijaya Bhaskar,
Research Associate,
Aurigene Discovery Technologies Limited,
Bioanalytical Division,
Bangalore,
Email: vijayb.aaa.aurigene.com
Back to the Top
Dear Dr. Tahir Kasumov,
I hope the following article will answer your query as I found it very
useful in choosing a weghting factor for calibration curves.
Almeida AM et al. "Linear regression for calibration lines revisited:
weighting schemes for bioanalytical methods". J Chrom B. 774(2002):
215 - 222.
regards
Akshanth
Back to the Top
The following message was posted to: PharmPK
Both weightings will provide pretty much the same result. One uses
weight based on the independent variable [conc, x], the other bases the
weighting on the response (y). Each will minimize the contribution of
high standards and maximize the contribution of the low standards to the
curve and curve fit. Other types of weighting (log) will adjust for the
high end of the curve and minimize the contribution of the low standards
The selection should be driven by using the simplest model. Linear
unweighted > linear weighted, Linear > quadratic > 4PL. The gauge to
measure models is the % Bias for each of the points.
--
Ed O'Connor, Ph.D.
Laboratory Director
Matrix BioAnalytical Laboratories
25 Science Park at Yale
New Haven, CT 06511
Web: www.matrixbioanalytical.com
Email: eoconnor.-a-.matrixbioanalytical.com
Want to post a follow-up message on this topic?
If this link does not work with your browser send a follow-up message to PharmPK@boomer.org with "Weighted calibration curves" as the subject | Support PharmPK by using the |
Copyright 1995-2011 David W. A. Bourne (david@boomer.org)