<tabu>

Description

A tabulation summarizes data by grouping and aggregating related data. For instance, if a tabulation summarizes the average temperature by city for a whole year, the city is the group and the temperature is the metric being aggregated by averaging all the temperatures on record for a given year.

The result of a tabulation is a table with columns for each break column specified by the breaks attribute and one result column for each aggregation specified with an individual <tcol> element. (See <tcol> for details.)

In the resultant table, the break columns retain their original names. To provide a different name for a break column or specify how it is to be sorted, optional <break> elements can be inserted between the opening and closing tags of the <tabu> element. See <break> for more information.

By default, the result columns are named t0, t1, etc. However, the result columns can be named using the name attribute in each of the <tcol> elements.

The cbreaks attribute specifies the columns for a cross-tabulation. If cbreaks is specified, the result is a table with one column for each row-break column (defined by the breaks attribute), another column containing row totals, and additional columns for each unique combination of column-break values.

In the resultant table, the row-break columns retain their original names (though they can be changed using the <break> element), the total column is called t0, and the result columns are named m0, m1, etc. (In the event of name collisions, special rules apply.) A <tcol> tag defines the type of aggregation.

As of version 10.17, cross tabulations may contain multiple <tcol> elements, allowing for multi-metric cross tabulations.

The columns specified by the multiple <tcol> elements are grouped together under each column break, and the row-total columns appear at the beginning in <tcol> order.

For instance, if there are m <tcol> elements and n values in the column breaks, there will be a total of m*(1+n) columns, which will be named: t0, t1, ..., tm, m0_0, m1_0, ..., mm_0, m0_1, m1_1, ..., mm_1, ...,m0_n, m1_n, ..., mm_n. (This is an extension of the regular naming scheme for cross-tabulations.)

Syntax

<tabu label="[TITLE_OF_TABULATION]" 
       breaks="[COLUMN(S)_TO_GROUP_BY]">
           <tcol source="[COLUMN(S)_TO_SUMMARIZE]" 
              fun="[NAME_OF_FUNCTIONS]"
              label="[NAME_OF_NEW_COLUMN]"/>
</tabu>

Attributes

label

Accepts a text value that specifies the title for the results table.  If omitted, the system supplies a default value.

breaks

A list of the names of the grouping ("break") columns (real or computed), separated by commas.  All records that have the same values for these columns will be grouped together.  For cross tabs, breaks specifies the row breaks only.  If omitted, all rows are treated as being in one group.

You can use an expression in the breaks attribute.

cbreaks

A list of names of the column-break columns (real or computed), separated by commas.  All records that have the same values for these columns will be grouped together.  If omitted, the result is a straight tabulation.

You can use an expression in the cbreaks attribute.

clabels

Accepts either the value long or the value short, indicating, for cross tabulations, whether the column headings of the result columns should include the grouping column names. 

If clabels="long", the column headings of the result columns will include the grouping column names and look something like Color=red Size=large.  If clabels="short", only the grouping columns' values will be used (e.g., red large).

xtab

The expression xtab=0 indicates that a cross-tabulation should be "unrolled." This means that for each group in breaks, there will be one row for each group in cbreaks.  The result is thus displayed in the form of a regular tabulation, with as many columns as there are <tcol> elements, but with one row for each possible combination of breaks and cbreaks. For combinations that do not occur, the result is NA, just as in a cross-tabulation.