Related Experiment Videos
Bioanalytical calibration curves: variability of optimal powers between and within analytical methods
1Department of Biometrics and Pharmacokinetics R & D, Phoenix International Life Sciences, Montreal, Quebec, Canada.
Journal of Pharmaceutical and Biomedical Analysis
|April 21, 1998
Summary
This study reveals that while analytical techniques vary, optimal power models show stable transformation parameters. Within-lab factors, not just analytical methods, likely cause variability and outlier rejection in calibration curves.
Area of Science:
- Analytical Chemistry
- Data Analysis
Background:
- Calibration curves are essential for quantitative analysis.
- Optimal power models are increasingly used but their variability is not well understood.
- Understanding outlier rejection frequency is crucial for data integrity.
Purpose of the Study:
- To investigate the variability of optimal power models versus common regression models.
- To assess variability within and between different analytical methods.
- To determine the frequency and patterns of outlier rejection in calibration data.
Main Methods:
- Fitting power models to calibration curve data using minimum sum of squared residuals for curve selection.
- Employing jackknife percent deviation for outlier detection.
- Analyzing data from 2087 analytical batches across 91 projects and various techniques.
Main Results:
- The most frequent regression model differed across analytical techniques.
- Median and interquartile range of optimal powers remained stable.
- Outlier rejection was highest in Gas Chromatography (GC) and Liquid Chromatography-Mass Spectrometry (LCMS), with the Wagner (Quadratic, log-log) model being most frequent.
Conclusions:
- Analytical technique is not the primary source of variability in optimal power transformations.
- Other within-laboratory factors likely contribute significantly to variability and outlier detection.
- Outlying values may stem from these unexamined sources of variability.