Model order selection results using F test. 

Contexts in source publication

Context 1

... breath-by-breath values of 2 O V & are sampled with a variable sampling period as the respiratory frequency changes due to breath-to-breath irregularities, and also, due to the need for increasing the ] T N t t , , 0 L = t (see Fig. 1 The estimation of o θ (see Eq. 2) at fixed order o from 2 vo is done classically by procedures that minimise the RSS between the experimental data points and the model [3, 4, 5, 6 & 11] i.e. the procedures are searching for the set of parameters that minimises the Euclidean norm of e o (see Eq. 3). This minimisation must be done with the following physiological constraints on parameters: ...

Context 2

... this questions the validity of another statistical approach used in previous physiological studies for the same goal [3,11]. This method retains the second-order model if the decrease in RSS is sufficient to offset the loss in degrees of freedom associated with the increased number of model parameters as determined by an F test. Table 2 shows the percentages of correct model order selection against both CV% and the order of the reference model, using this method. It is well apparent that it reacts like Information Criteria incorporating small β values and penalisation terms: comparable results are obtained with 0 < β < 0.1. ...