Precision and Accuracy of AH Measurements

The variables of heating rate, sample size, instrument sensitivity level, and metal standard used were incorporated into a factorial, 3 x 23 design. To keep the number of runs within reasonable limits, Schwenker and Whitwell used only three metals, indium, tin, and lead, as standards. 

The predicted influence of the number of peak area measurements on calibration coefficient precision is shown in Table 5.10 (114). These data indicate that the best procedure is to replicate samples rather than plani-meter area measurements. The standard errors show that no significant improvement in precision results from making four area measurements instead of two. whereas precision is markedly improved by increasing the number of samples. 

The effect of the number of samples used to determine K and the number of samples for determining the AH of an unknown transition on precision, at the 95% confidence limits, is shown in Table 5.11. These results indicate that for the higher levels of precision, several samples are required for calibration. 

From a practical viewpoint, very few calorimetric studies are reported in which 30 samples are used to calibrate K and to determine AH values. Hence, the precision expected is certainly greater than 1%, as indicated in Table 5.11. Since most of the measurements are of ihe first two types, the precision expected is from 4-7%, if not greater. 

Brennan and Gray 1156) discussed the calorimetric precision and accuracy of DSC measurements (the Perkin-Elmer DSC-2). They stated that the operating principle and the design of the DSC-2 instrument provides no fundamental reason for it to be limited to a precision of I -2%, as commonly reported. They examined the effects of sample mass, heating rate, and instrument sensitivity on the calibration of the instrument by a factorial design similar to that of Schvvenker and Whitwell 114). It was found that no syste- 

---