Error estimates on normal least squares linear regression with replicate injection of calibration standards

Area of Science:

• Analytical Chemistry
• Bioanalysis
• Method Validation

Background:

• Normal least squares linear regression models are used in bioanalysis.
• Calibration standards can be analyzed via single or replicate injections.
• Different regression models can be generated based on data handling.

Purpose of the Study:

• To evaluate the impact of different calibration standard data handling methods on bioanalytical results.
• To compare the precision and confidence intervals of sample results derived from various linear regression models.

Main Methods:

• Generation of four normal least squares linear regression models using different calibration standard datasets (single injection, averaged injections, all injections).
• Analysis of sample results and their estimated confidence intervals from each model.

Main Results:

• Models using averaged (C) or all (D) injection data yield identical slopes and intercepts.
• Models C and D produce narrower confidence intervals for sample results compared to models A and B.
• This improvement is attributed to reduced standard error and increased calibration points, leading to a lower Student's t-value.

Conclusions:

• Replicate injection of calibration standards enhances analytical measurement precision.
• This approach offers benefits in instrument sensitivity compensation.
• Utilizing all available calibration data improves the reliability of bioanalytical results.