Measures of variability of the least-squares estimator theta are essen
tial to assess the quality of the estimation. In nonlinear regression,
an accurate approximation of the covariance matrix of theta is diffic
ult to obtain [4]. In this paper, a second-order approximation of the
entropy of the distribution of theta is proposed, which is only slight
ly more complicated than the widely used bias approximation of Box [3]
. It is based on the ''flat'' or ''saddle-point approximation'' of the
density of theta. The neglected terms are of order O(sigma4), while t
he classical first order approximation neglects terms of order O(sigma
2). Various illustrative examples are presented, including the use of
the approximate entropy as a criterion for experimental design.