Multivariate tests controlled for demographics, eliminated redundant variance, identified which factors were most strongly associated with progression, identified which stages were most influenced by each predictor, and indicated the direction of each predictor’s effect after accounting for all other variables (see Table 3). In all three time periods, 8 to 12 independent predictors explained significant variation in the survival curve, and the pattern of effects within each period was distinct from the pattern of effects in the other periods. Use of z-scores allowed straightforward comparison and interpretation. In general, results indicated that the multivariate pattern of findings was consistent with the univariate pattern.
