When estimating the unknown parameters in semi-empirical relationships from vapour-liquid equilibrium experiments, it has been common practice to minimize the sum of squares of some arbitrary function of one or more of the measured variables (x, y, P, T) without regard to the statistical consequences of such a procedure. It is shown that because these arbitrary procedures do not fully account for the errors in all the variables in the functional relationship they can lead to poor parameter estimates when compared to a statistically sound procedure based on maximum likelihood. The analysis of some binary data sets and a simulation study based on one of them are used to demonstrate the kinds of errors that arise.
Use of the maximum likelihood method of estimation also aids one in checking the adequacy of the model used, and in testing the consistency of the experimental data. This is accomplished by comparing the deviations between predicted and measured responses with their computed confidence limits.
En evaluant les parametres inconnus dans des relations semi-empiriques etablies apartir d'experiences sur l'equilibre entre vapeur et liquide, on minimise generalement la somme des carres d'une fonction arbitraire de l'une ou plus des variables mesurees (x, y, P, T), sans tenir compte des consequences statistiques de cette maniere de proceder. On constate que lesdits procedes arbitraires, qui ne tiennent pas completement compte des erreurs dans toutes les variables de la relation fonctionnelle, peuvent conduire ades evaluations mediocres des parametres, si on les compare aune methode statistiquement valable qui est basee sur la probabilite maximale. L'analyse de certains resultats et une etude de simulation basee sur l'un d'entre eux servent ademontrer les genres d'erreurs qui se produisent.
L'emploi de la methode d'evaluation dite de probabilite maximale aide aussi a verifier l'efficacite du modele ainsi que la stabilite des resultats experimentaux; on y parvient en comparant les ecarts qui existent entre les reponses prevues et mesurees avec leurs limites calculees de credibilite.
[1]
C. Kretschmer,et al.
Densities and liquid-vapor equilibria of the system ethanol-isoöctane (2,2,4-trimethylpentane) between O and 50 degrees.
,
1948,
Journal of the American Chemical Society.
[2]
F. Stevenson,et al.
Effects of Experimental Errors on Thermodynamic Consistency and on Representation of Vapor-Liquid Equilibrium Data
,
1972
.
[3]
M. J. Box.
Improved Parameter Estimation
,
1970
.
[4]
C. Tsonopoulos,et al.
An empirical correlation of second virial coefficients
,
1974
.
[5]
L. Verhoeye.
Remarks on the determination of the Wilson constants in the correlation of vapour—liquid equilibrium data
,
1970
.
[6]
E. Neau,et al.
Thermodynamics of methanol-1,2 dichloroethane system
,
1976
.
[7]
Andre Peneloux,et al.
The maximum likelyhood test and the estimation of experimental inaccuracies. Application to data reduction for liquid-vapor equilibrium
,
1976
.
[8]
Jean-Francois Fabries,et al.
Method of evaluation and reduction of vapor‐liquid equilibrium data of binary mixtures
,
1975
.
[9]
Mitsuho Hirata,et al.
ESTIMATION OF WBLSON PARAMETERS
,
1971
.
[10]
N. Brinkman,et al.
A study of the effect of optimization function on calculated parameters of wilson equation
,
1974
.
[11]
John F. MacGregor,et al.
ASQC Chemical Division Technical Conference 1971 Prize Winning Paper Some Problems Associated with the Analysis of Multiresponse Data
,
1973
.