rowRank

Syntax

rowRank(X, [ascending=true], [groupNum], [ignoreNA=true], [tiesMethod='min'], [percent=false], [precision])

Please see rowFunctions for the parameters and calculation rules.

Parameters

X is a matrix.

ascending (optional) is a Boolean value indicating whether the sorting is in ascending order. The default value is true (ascending).

groupNum (optional) is a positive integer indicating the number of groups to sort X into.

ignoreNA (optional) is a Boolean value indicating whether null values are ignored.

tiesMethod (optional) is a string indicating how to rank the group of elements with the same value (i.e., ties):

  • 'min' : the smallest rank value of the tie values.

  • 'max' : the largest rank value of the tie values.

  • 'average' : the average of the rank values for all ties.

  • 'first': Gives the first found tie value the lowest rank value, and continues with the following rank value for the next tie.

percent (optional) is a Boolean value, indicating whether to display the returned rankings in percentile form. The default value is false.

precision (optional) is an integer between [1, 15]. If the absolute difference between two values is no greater than 10^(-precision), the two values are considered to be equal.
Note:

If parameter precision is specified, X must be numeric, and the tiesMethod cannot be specified as 'first'.

Details

Conduct the following operation within each row of matrix X:
  • Return the position of each element in the sorted vector.
  • If groupNum is specified, group the elements into groupNum groups and return the group number each element belongs to.
  • If ignoreNA =true, null values return NULL.

The result is a matrix with the same shape as X.

Examples

m=matrix(3 1 2 4 7 6 9 8 5, 9 NULL 2 3 5 6 3 2 8).transpose();
m
#0 #1 #2 #3 #4 #5 #6 #7 #8
3 1 2 4 7 6 9 8 5
9 2 3 5 6 3 2 8 
m.rowRank();
#0 #1 #2 #3 #4 #5 #6 #7 #8
2 0 1 3 6 5 8 7 4
7 0 2 4 5 2 0 6 
m.rowRank(ascending=false);
#0 #1 #2 #3 #4 #5 #6 #7 #8
6 8 7 5 2 3 0 1 4
0 6 4 3 2 4 6 1
Specify percent=true to return the ranking in fractional (percentage) form.
m.rowRank(percent=true)
0 1 2 3 4 5 6 7 8
0.33 0.11 0.22 0.44 0.77 0.67 1 0.89 0.56
1 0.125 0.375 0.625 0.75 0.375 0.125 0.875 
m.rowRank(groupNum=3);
#0 #1 #2 #3 #4 #5 #6 #7 #8
0 0 0 1 2 1 2 2 1
2 0 0 1 1 0 0 2 
m.rowRank(ignoreNA=false);
#0 #1 #2 #3 #4 #5 #6 #7 #8
2 0 1 3 6 5 8 7 4
8 0 1 3 5 6 3 1 7 
m.rowRank(ignoreNA=false, tiesMethod='max');
#0 #1 #2 #3 #4 #5 #6 #7 #8
2 0 1 3 6 5 8 7 4
8 0 2 4 5 6 4 2 7 
m.rowRank(ignoreNA=false, tiesMethod='first');
col1 col2 col3 col4 col5 col6 col7 col8 col9
2 0 1 3 6 5 8 7 4
8 0 1 3 5 6 4 2 7
For floating-point numbers, the comparison precision can be controlled using the precision parameter. When the absolute difference between two values is no greater than 10^(-precision), they are considered equal. In the example below, when the precision parameter is not specified, 1.000003 and 1.000004 are treated as unequal by default.
m = matrix(1.000001 1.000002 1.000003,   
           2.000001 2.000002 2.000003,  
           1.000001 1.000002 1.000004) 
rowRank(m);
col1 col2 col3
0 2 0
0 2 0
0 2 1
After setting precision = 4, 1.000003 and 1.000004 are considered equal (both receive a rank of 0).
rowRank(m, precision=4)
col1 col2 col3
0 2 0
0 2 0
0 2 0

Related functions: rowDenseRank, rank