From this raw importance score a standardized importance score can be computed with the following rationale: The individual importance scores VI (t)(xj) are computed from ntree bootstrap samples, that are independent given the original sample, and are identically distributed. Thus, if each individual variable importance VI (t) has standard deviation σ, the average importance from ntree replications has standard error σ/ntree. The standardized or scaled importance, also called “z-score”, is then computed as
