Correlation Coefficient Analysis of Polygraphic Variables
We subjected all data to Pearson correlation coefficient analysis. As this analysis has limitations, particularly when many variables are concerned, we have indicated only the correlation coefficient. We found correlations between mean Sa02 and percent TOO percent Sa02 (0.61), between percent TOO percent Sa02 and 02-90-I (0.88), and between lowest Sa02 and 02-80-I (0.58). Because 02-90-I was so closely correlated to percent TOO percent Sa02, it was discarded from further analysis. We decided, instead, to select percent TOO percent Sa02 and 02-80-I as Sa02 indices.
These two indices were also correlated (0.60), but they gave two types of information pertinent to OS AS: an overall view of oxygenation during sleep vs a notion of repetitive drops of Sa02. There was also a positive correlation between short latencies at night and at MSLT (0.82), and between short nocturnal REM sleep latency and daytime REM sleep latency (0.74). Several other variables were positively correlated, particularly the amount of stage 3-4 NREM sleep (S3-4) and long sleep latency at MSLT (0.66); long nocturnal TST and percentage of stage 2 NREM sleep (S2) (0.73); and percentage of SI and amount of wake time during nocturnal sleep (0.46). There were negative correlations between age and percentage of NREM sleep ( — 0.48); RDI and amount of nocturnal S3-4 ( — 0.49); RDI and amount of nocturnal S2 ( — 0.58); and long sleep latency at MSLT and amount of nocturnal SI ( — 0.51). The positive and negative correlations indicated that the nocturnal sleep disturbances were reflected in the daytime mean sleep latency results. Because some of the many variables correlated could be significant only because of the large number considered, we chose to analyze our data more rigorously.