*3.4.2 Precision*

Standard deviation (s) and the coefficient of variation (CV) must be determined in order to analyze the degree of agreement between the individual data obtained when the method is repeatedly applied to multiple aliquots of a homogeneous sample.

Accuracy is evaluated at two levels: Repeatability and Intermediate accuracy as described below.

### *3.4.2.1 Repeatability*

We analyzed the precision obtained after performing five calibration curves, in a concentration range of 0.1 to 1.0 μg/L for cadmium, where absorbance is measured four times per standard, under the same operating conditions in a short time interval (same day) by the same analyst and using the same equipment, materials, and reagents. This value corresponds to 0.136 [54, 70].


**Table 3.**

*Regression analysis for model Y = β0+ β1x.*

#### *3.4.2.2 Intermediate precision*

We analyzed the precision obtained after performing seven calibration curves, in the same concentration range as for repeatability, where absorbance is measured four times per standard under the same operating conditions, in different time intervals (7 different days), by the same analyst and using the same equipment, materials, and reagents. This value corresponds to 0.135 [54, 70].

This method increases sensitivity when the analysis is conducted on the same day, indicating that the proposed method for sample preparation is appropriate. The sensitivity measured for the method was 0.136.

For the two previous procedures (Repeatability and intermediate precision), the Shapiro–Wilk test was applied with N-1 degrees of freedom (N = readings of absorbance) and Homogeneity of variances with degrees of freedom 1 and 2 calculated as K-1 and (k-N)-K respectively (where N = readings of absorbance and K = number of calibration curves), posing the corresponding null and alternative hypotheses at a 95% confidence level. In addition, for each of the (standard) concentration levels, the respective standard deviations and variation coefficients were obtained to determine whether the method is accurate.

For repeatability, the results show that at a 95% confidence level, the calculated W is lower than the tabulated W (0.999) for all levels of metal concentration, with 3 degrees of freedom for cadmium. Thus, the null hypothesis is accepted and it is concluded that the data come from a normal distribution. Applying Levene's test for cadmium data, we can see that for 4 and 15 degrees of freedom, the calculated statistic is lower than the tabulated statistic for all levels of concentration. Thus, we conclude that the variances are homogeneous. The standards present a low standard deviation (less than 0.004) and a coefficient of variation that is lower or equal to 5%, thus indicating that the method used presents a good repeatability [54, 70].

For the intermediate precision, the calculated Shapiro–Wilk statistic was lower than the tabulated statistic (0.999), therefore concluding that the data come from a normal distribution at a 95% confidence level. Taking into account Levene's statistic, the calculated statistic was lower than the tabulated one, thus concluding that the variances are homogeneous [54, 70].

The metal standards, show a small deviation (less than 0.005) and a coefficient of variation that is lower than or equal to 4.0%, leading to the conclusion that the method is of good accuracy [54, 70].

#### *3.4.3 Sensitivity*

Sensitivity is assessed as analytical sensitivity and calibration sensitivity, as described below.

#### *3.4.3.1 Analytical sensitivity*

The parameters used to determine analytical sensitivity are the limit of detection and quantification. Fifteen absorbance readings were taken from the metal target and the standard deviation calculated along with the detection and quantification limits, following the method suggested by IUPAC (1995) [70, 71].

The LOD of Cd was 0.02 mg/L (0.4 mg/kg) and corresponds to the minimum amount of Cd derived from the lowest analytical signal that can be detected with reasonable certainty. The LOQ of Cd was 0.07 mg/L (1.4 mg/kg) and represents the minimum concentration that can be measured with precision and accuracy. The LOD and LOQ are adequate for the quality control of biopolymers [54, 70].
