As an example, the solubility of naphthalene in supercritical CO2 and C2H4 were chosen. Two parameter estimation criteria were compared : use of constant relative error and the maximum likelihood criterion, the latter involves the use of weights derived from the error propagation law. It was found that sigma(DELTAy)2 is very far from being proportional to y2, therefore the use of constant relative error is not justified. In the case of the given solid-fluid equilibrium examples, weights based on constant relative error assumption overemphasize the steep part of the curve where all models fail. None of the models examined were able to describe C2H4-naphthalene equilibrium. For CO2-naphthalene the BMRK model proved to be better than the other equations of state investigated. : use of constant relative error and the maximum likelihood criterion, the latter involves the use of weights derived from the error propagation law. It was found that sigma(DELTAy)2 is very far from being proportional to y2, therefore the use of constant relative error is not justified. In the case of the given solid-fluid equilibrium examples, weights based on constant relative error assumption overemphasize the steep part of the curve where all models fail. None of the models examined were able to describe C2H4-naphthalene equilibrium. For CO2-naphthalene the BMRK model proved to be better than the other equations of state investigated.