Basic Principles of Measurement
Sarah Armstrong, Barry Clifton, Lionel Davis in Primary FRCA in a Box, 2019
Calibration is the setting up or correction of a measuring device or base level, usually by adjusting it to conform to a dependably known and unvarying measureA calibration curve is the graphical representation of the functioning relationship between the expected value of the observed signal to the measured amountCalibration aims to remove the effects of drift on the measurement. Drift may be gradient drift (measurement value increases disproportionately to increasing input value) or offset drift (where every measured value has a constant drift). These require one-point calibration. A combination of offset and gradient drift requires two-point calibration
Does my model predict accurately?
Thomas A. Gerds, Michael W. Kattan in Medical Risk Prediction, 2021
A model is well-calibrated (in the large) if 17% of the subjects have the event when the predicted risk for all of them is 17%. The value 17% is just an example. This should hold for all values that the predicted risk can take, i.e., all values between 0% and 100%. The calibration curve is defined over the range of values that the model is possibly predicting. For any such value the calibration is the observed event frequency among patients that have a predicted risk equal or at least close to that value. There are many different ways to define “close,” and a popular method is to simply categorize the predicted risks according to deciles [123]. This yields the histogram type of the calibration plot. Figure 5.7 shows a calibration diagram based on 10 groups of the predicted risks that are obtained using deciles. The model is well-calibrated (in the small) if the light gray bars are as high as the dark gray bars.
C
Filomena Pereira-Maxwell in Medical Statistics, 2018
A procedure by which measurements that were obtained by means of a method or tool that might be biased are compared against an accepted standard. A correction factor for the method or tool in question is derived from such an exercise, often by means of regression modelling. A calibration curve may be drawn, allowing the estimated true value for the measurement to be read off, given the value obtained with the biased method. See also accuracy, measurement bias, method comparison studies. The term is also used to refer to an assessment of the accuracy of a predictive model, as is the case when the Hosmer and Lemeshow goodness-of-fit statistic is computed to evaluate the predictive accuracy of logistic regression models. Cf. discrimination.
Novel model predicts diastolic cardiac dysfunction in type 2 diabetes
Published in Annals of Medicine, 2023
Mingyu Hao, Xiaohong Huang, Xueting Liu, Xiaokang Fang, Haiyan Li, Lingbo Lv, Liming Zhou, Tiecheng Guo, Dewen Yan
The calibration curve is the consistency between the frequency of observed results and prediction probability. Research calibration is expressed by following the relationship between the frequency of the effect and the predicted probability. A sensible calibration measure is a likelihood ratio statistic testing the null hypothesis that intercept = 0 and slope = 1. The statistic has a χ2 distribution with 2 degrees of freedom (unreliability U-statistic) [20]. We also evaluated average (E-aver) and maximal errors (E-max) between predictions and observations obtained from a calibration curve. Plotted the calibration curve to assess the calibration of the nomogram, and the nonsignificant test statistics show that the model has been perfectly calibrated [21]. Decision curve analysis (DCA) was used to evaluate the clinical value of the predictive model. Decision curve analysis was conducted to determine the clinical usefulness of the nomogram by quantifying the net benefits at different threshold probabilities in the validation dataset [22].
UV–Vis spectroscopic quantification of residual acetone during the development of nanoparticulate drug delivery systems
Published in Pharmaceutical Development and Technology, 2019
Sergio M. Espinoza, Rocio Guadalupe Casañas Pimentel, Eduardo San Martin Martinez
With optimum conditions defined, validation started with the assessment of linearity. Acetone concentrations in the range of 10–50 µg/mL were quantified employing the already found optimum conditions, given as result a linear plot (Figure 3(b)) with an R2 of 0.998. The calibration curve is described below as Equation (4): y corresponds to absorbance and x to acetone concentration in µg/mL. The limit of detection (LOD) and limit of quantification (LOQ) were quantified as 3.3 and 10 times the residual standard deviation (SD) of the regression divided by the regression slope, with the obtained values corresponding to 2.6 and 7.8 µg/mL for LOD and LOQ, respectively. Precision, on the other hand, was assessed as repeatability and intermediate precision (inter-day precision). Data showed a relative standard deviation (RSD) better than 5.8% and as good as 1.5%, resulting satisfactory for the intended application of the method. Further assessment of days effect showed no significant difference (p = 0.1530), implying that the method could be successfully applied regardless of the day it is performed and presenting suitable intermediate precision in this regard. As for accuracy, it was evaluated as recovery percent from acetone solutions of known concentration (40, 30, and 20 µg/mL). Results showed satisfactory accuracy at the three levels, with a 103, 107, and 115% recovery percentages for high, medium, and low acetone concentrations, respectively.
New insights in the metabolism of oxybutynin: evidence of N-oxidation of propargylamine moiety and rearrangement to enaminoketone
Published in Xenobiotica, 2018
Silvio Aprile, Rossana Canavesi, Rosanna Matucci, Cristina Bellucci, Erika Del Grosso, Giorgio Grosa
Stock solutions of oxybutynin, Oxy-DE, Oxy-EK, and Oxy-HA were prepared in methanol at 1 mg/ml concentration. Standard working solutions were freshly obtained by serial dilution of the stock solutions with methanol. The calibration standards were prepared by spiking appropriate aliquots of working solutions (5 μl) into blank plasma or urine (45 μl) and diluting with 100 or 50 μl of ice-cold acetonitrile respectively. Calibrators were homogenized and centrifuged at 12,500 rpm for 10 min. Supernatants (10 μl) were injected into the LC–MS system. The calibration curves (y = ax + b) were constructed from the peak areas versus analyte concentration (over a range of 5–1000 μg/l) using the weighted (1/x) linear least-squares regression method. Precision and accuracy were determined by repeated analyses (n = 3) of the five calibrators used to construct the calibration curve for each analyte. Precision was expressed as percent coefficient of variation (%CV), whereas the accuracy was determined as the percent relative error (%RE) of the mean back calculated concentrations from the theoretical concentrations. The mean precision (%CV) for all the analytes was ≤20%, whereas the mean accuracy (%RE) were 11.1%, 18.1%, 4.1%, and 17.8% for Oxybutynin, Oxy-DE, Oxy-EK and Oxy-HA respectively.
Related Knowledge Centers
- Bradford Protein Assay
- Chromium
- Coomassie Brilliant Blue
- Reagent
- Analytical Chemistry
- Internal Standard
- Analyte
- Instrumentation
- Sensor
- Photomultiplier Tube