Two calculation methods for estimating reactivity ratios, one method based on the differential Alfrey‐Mayo equation and one based on the integrated form of this model, are compared with respect to precision and bias. Both methods are characterized by the use of information about the monomer feed composition only and are assumed to be valid up to high conversion. As only monomer feed composition has to be analyzed, several sampling designs are feasible. Two extreme designs can be distinguished. One consists of repetitive sampling of the initial and final monomer feed mixture, whereas the other consists of sequential sampling during the course of the reaction. The influence of both designs of the calculated r‐values is investigated by means of simulation. In the present paper the second calculation method, based on the integrated form, is solved by a nonlinear least squares method considering errors in both variables. This method required additional information about the errorstructure of the data. As this information is mostly of approximate nature, the influence of misspecification of this error structure on the calculated r‐values is also examined.

SynopsisFive calculation methods for estimating reactivity ratios are compared. Two of these methods are only valid at low conversion, two methods are assumed to be valid up to high conversion, and the fifth method is the integrated form of the copolymer equation. Data were simulated for selected couples of r-values and two conversion levels. The data were randomly disturbed by normal error with mean zero. As (the) monomer feed ratids) will be drifting during the course of reaction, the influence of a n approximated monomer feed ratio on the r-values calculated was also examined. When conversion is low, all methods give estimates with low precision. High conversion results in larger precision, however, for several methods bias a g pears. For calculation methods that need information about an approximated monomer feed ratio, the influence of this approximation appears to be rather important especially if the rvalues are dissimilar or conversion is high.

No abstract