Table 4. Sign of Aggression-Change on Statins vs Placebo in Men.

(Sign of change is insensitive to outliers.)

	OR (SE)	P	95%CI
Statins N = 664	0.64 (0.097)	0.003	0.47, 0.86
Simvastatin N = 443	0.52 (0.094)	<0.001	0.37, 0.75
Pravastatin N = 441	0.77 (0.14)	0.14	0.54, 1.1

CI = confidence interval; OASMa = Overt-Aggression-Scale-Modified–Aggression-Subscale; OR = odds ratio from ordinal logit; SE = standard error. Analysis employs ordinal logit with robust standard errors, not adjusted for baseline OASMa. There is no exclusion of outliers, as no values represent outliers (values are collapsed to -1, 0, +1—reflecting reduction, no change, or increase in aggression, respectively). This analysis is insensitive to large magnitude outliers, as it looks only at direction (sign) not magnitude. For odds ratios derived from ordinal logit, “a unit change in the predictor variable signifies that the odds for the outcome being in a group that is greater than k versus less than or equal to k is the proportional odds times larger” [55].