SPSS Rank
The RANK command sets a variable equal to the case rank.
Contents
Usage
rank variables=foo bar(A) baz(D). rank variables=foo bar(A) baz(D) /rank into new1 new2 new3.
See SORT CASES for the syntax of ranked variables.
The /RANK subcommand is the default action. Specifying that subcommand allows for specifying output variable names. If the subcommand is not specified, or if it is specified without any or with too few output variable names, unique variable names will be automatically chosen.
Ties
Strategies for handling ties are:
Option |
Effect |
MEAN |
Mean of tied ranks |
LOW |
Lowest of tied ranks |
HIGH |
Highest of tied ranks |
CONDENSE |
Consecutive ranks for all groups |
The default is /TIES=MEAN
foo |
rank_mean |
rank_low |
rank_high |
rank_condense |
0 |
3 |
1 |
5 |
1 |
0 |
3 |
1 |
5 |
1 |
0 |
3 |
1 |
5 |
1 |
0 |
3 |
1 |
5 |
1 |
0 |
3 |
1 |
5 |
1 |
1 |
6.5 |
6 |
7 |
2 |
1 |
6.5 |
6 |
7 |
2 |
2 |
8 |
8 |
8 |
3 |
Ntiles
Compute the percentile for each N-tile with the /NTILES subcommand. For example, /NTILES(4) causes each quartile to be computed.
rank variables=foo bar baz /ntiles(4) into new1 new2 new3.
Proportion
Compute the proportion estimate with the /PROPORTION subcommand.
rank variables=foo bar baz /fraction=blom /proportion into new1 new2 new3.
The /FRACTION subcommand specifies the method for computing a proportion estimate. By default, SPSS will use Blom's transformation. Alternatives include:
Method |
Effect |
BLOM |
Blom's transformation: (r – 3/8) / (w + 1/4) |
RANKIT |
(r – 1/2) / w |
TUKEY |
Tukey's transformation: (r – 1/3) / (w + 1/3) |
VW |
Van der Waerden's transformation: r / (w +1) |
Normal
Compute the inverse of the standard normal cumulative distribution of the proportion estimate with the /NORMAL subcommand.
rank variables=foo bar baz /fraction=blom /normal into new1 new2 new3.
The /FRACTION subcommand specifies the method for computing a proportion estimate. See /NORMAL for the specification.
N
Compute the number of cases, or sum of weights, with the /N subcommand.
rank variables=foo bar baz /n into new1 new2 new3.
RFraction
Compute the fractional ranks with the /RFRACTION subcommand. The computed values are ranks divided by the number of cases, or the sum of weights.
rank variables=foo bar baz /rfraction into new1 new2 new3.
Combined with /TIES=HIGH, this is equivalent to cumulative frequencies.
Percent
Compute the fractional ranks as a percentage with the /PERCENT subcommand. The computed values are as created by /RFRACTION multiplied by 100.
rank variables=foo bar baz /percent into new1 new2 new3.
Savage
Compute the Savage (exponential) scores with the /SAVAGE subcommand.
rank variables=foo bar baz /savage into new1 new2 new3.
By
To rank cases within groups, specify the BY option.
rank variables foo bar by baz.
Missing
By default, SPSS will exclude observations with a user missing value for any variable. Add the /MISSING=EXCLUDE subcommand to explicitly allow this behavior, or add /MISSING=INCLUDE subcommand to disable it.
Data model
The RANK command executes immediately. If there are any pending transformations, they execute first.