Design and analysis of progress curves in enzyme kinetics.

Lastly, I wish to illustrate and to stress the necessity of assessing goodness of fit by making use of several tests and not relying on only one or two such tests. Referring to Table 1, the xz test and the Runs test indicate that model 3 is a good fit, whereas the curved Rankits plot, the slow convergence, the non-linearity at the minimum sum of squares and the wide range of the coefficients of variation for the parameters cast doubt on the fit. The model must, however, be rejected because one of the terms in the model contributes little to the model. For model 1, the parameter values and their S.D. values, the convergence and the xz test indicate a good fit. However, the Rankits plot and the sum-of-squares surface suggests some doubt and the Runs test especially means that the model is very doubtful. The tests on model 2 are all satisfactory apart from the Runs test. Hence by assessing a wide variety of evidence one can conclude that Model 2 is a good model and the other two are in several ways less satisfactory. In conclusion: (1) before it can be accepted that a given model is a good description of a particular system, sufficient statistical evidence must be presented ; (2) one or two tests on the fit of a model are often insufficient; as many tests as possible must be used.