On the maximum likelihood estimation of respiratory response slopes

Several approaches have been suggested for estimating a respiratory response slope when both x and y variables are observed with error. Recently, a maximum likelihood estimate under the assumption of a bivariate normal distribution has been proposed. A method of moments solution yields a slope estimate of y/x as long as the underlying process mean is nonzero. This paper extends the maximum likelihood approach to the case where the process mean is zero. In this case, certain additional error assumptions must be made to yield a unique estimate. These concepts are applied to the problem of estimating an effective lung volume for steady-state breath-to-breath gas exchange data during exercise.