Prediction Accuracy in Multivariate Repeated-Measures Bayesian Forecasting Models with Examples Drawn from Research on Sleep and Circadian Rhythms

Figure 8

Cost of collection and number of nights of actigraphy and polysomnography (PSG) which combined will yield a certain average accuracy of wakefulness after sleep onset (WASO) parameter estimates. For illustration purposes, PSG and actigraphy are assumed to cost $1250 and $150 per night, respectively. The cost of data collection is fixed on each diagonal dashed line (for illustrative purposes these are only shown at fixed intervals) and increases as we collect more nights of PSG and actigraphy. Subject-specific WASO estimates of an average accuracy of (a) 15 min, (b) 14 min, and (c) 13 min are obtained on the solid curve. The point that minimizes the cost of obtaining the fixed accuracy (open squares) is determined as the point on the fixed accuracy curve where the line tangent to the curve is parallel to the fixed cost lines. For comparison the solution for obtaining the same average accuracy using only polysomnography is also shown (solid squares).

(a)

(b)

(c)