User Guide

145
Data Transform
ations
Rank Cases: Types
You can select multiple ranking methods. A separate ranking variable is created for
each method. Ranking methods include simple ranks, Savage scores, fractional
ranks, and p
ercentiles. You can also create rankings based on proportion estimates
and normal scores.
Rank. Simple rank. The value of the new variable equals its rank.
Savage scor
e.
The new variable contains Savage scoresbasedonanexponential
distribution.
Fractional rank. The value of the new variable equals rank divided by the sum of the
weights of
the nonmissing cases.
Fractional rank as percent. Each rank is divided by the number of cases with valid
values and multiplied by 100.
Sum of case
weights.
The value of the new variable equals the sum of case weights.
The new variable is a constant for all cases in the same group.
Ntiles. Ranks are based on percentile groups, with each group containing
approxima
tely the same number of cases. For example, 4 Ntiles would assign a
rank of 1 to cases below the 25th percentile, 2 to cases between the 25th and 50th
percentile, 3 to cases between the 50th and 75th percentile, and 4 to cases above
the 75th p
ercentile.
Proporti
on estimates.
Estimates of the cumulative proportion of the distribution
corresponding to a particular rank.
Normal scores. The z scores corresponding to the estimated cumulative proportion.
Proportion Estimation Formula. For proportion estimates and normal scores, you can
select the proportion estimation formula: Blom, Tukey , Rankit,orVan der Waerden.
 Blom. Creates new ranking variable based on proportion estimates that uses the
formula (r-3/8) / (w+1/4), where w is the sum of the case weights and r is the rank.
 Tukey. Uses the formula (r-1/3) / (w+1/3), where r is the rank and w is the sum
of the ca
se weights.