A procedure is described for fitting mathematical models to ventilatory response data resulting from a sequence of impulse stimuli. Since the breath-breath response data are time sequential, an autoregressive term is included in the model to remove serial correlation, thus helping to assure meaningful tests of significance. The model fitting technique used is separable least squares, which is a combination of non-linear (quasi-newton) and linear (least squares) methods, with a linear regression nested within the nonlinear interations. A 10 second impulse stimulus was generated by a computer controlled cycle ergometer capable of 'true zero watt' pedaling. True passive pedaling permits an increased stimulus amplitude without exceeding the anaerobic threshold, thus avoiding the response non-linearities that occur with lactic acid production. Each subject had two independent trials at a baseline workload of either 0 or 25 watts. After a two minute warmup at baseline, the subjects received five ten second impulses of 100 watts separated by seven minute recovery periods. The subjects maintained a constant pedaling rate of 60 r/min. The breath-breath minute ventilation, as determined from tidal volume and breath time, and end tidal gases were recorded digitally. The minute ventilation data were fitted to a first order model (single exponential), and a two component model with no exponential terms. The higher order model was a statistically significantly better fit in 3 of the 5 subjects.
|Original language||English (US)|
|Number of pages||8|
|Journal||Biomedical Sciences Instrumentation|
|Publication status||Published - 1982|
ASJC Scopus subject areas
- Hardware and Architecture