How can I calculate accuracy? 

A low coefficient of variation (CV = SD / mean) implies good accuracy. Always compare the variation in accuracy vs. the variation of the measured population.

Accuracy is a measure of how well an indirect method (such as the Mid-IR) is compared to a reference method (or against the “true value”). Analytical accuracy can be specified with standard statistical measurements. Two different examples: 

Differential Method
Calculation of the difference between HMA and the reference method. D = x – y 
Mean value of the differences, D = ƩD / n 
Standard deviation (SD) of the differences (D = x – y) between HMA (y) and reference (x) 
SD = Ʃ(D2 – (ƩD/n)2) / n – 1

Regression Method
Yi = a + bXi + ei 
Yi = dependent variable (reference method) 
a = intercept (or bias) 
b = slope 
Xi = independent variable (HMA) 
ei = variance of linearity (accuracy)

How can I calculate repeatability?

The intra-assay variation can be calculated in different ways. It is similar to precision or repeatability. Double measurements of n different samples gives the possibility to calculate the within variation using an analysis of variance (ANOVA) model. 

Between samples  
Within samples  

The estimated within variation gives the methods intra-assay variation (precision, repeatability). These statistical models will be found in most statistical books. You can also calculate the intra-assay variation by calculating the variance of difference between measurement 1 and 2. In ISO 8196-1:2000 a simplified equation is given how to calculate the standard deviation of the difference between double measurements of different samples.

Why doesn’t the instrument show exactly the same values every time? 

Milk is a very sensitive fluid and the value may change a little bit over time if you keep it standing in water bath. Milk is not a true solution, it is an emulsion (a mixture of liquids that normally does not mix easily, such as water and oil) and therefore every small sample drawn from a bottle will differ a little bit. You also have to keep in mind that there is a natural variation when analysing as there are a lot of factors that affect the result (both Mid-IR instrument and sample factors). As long as the result is within the specification you don’t have to worry about small differences.

What is the measurement range for each parameter?

Fat 0.6 - 5.9 [g/100ml]

Crude protein 0.8 - 3 [g/100ml]

True Protein 0.6 – 2.4 g/100 ml

Carbohydrate 4-8 [g/100ml]

How do I know if the instrument displays correct values?

Miris declares the following measurement performance:

Precision: fat ± 12%; crude protein, true protein, carbohydrate ± 15%  

Repeatability: fat, crude protein, true protein ≤0.05 g/100ml; carbohydrate ≤0.08 g/100ml

This is validated for each instrument individually during the production process.