Grouping Variables To Split Data

You can use grouping variables to split data variables into groups. Typically, selecting grouping variables is the first step in theSplit-Apply-Combineworkflow. You can split data into groups, apply a function to each group, and combine the results. You also can denote missing values in grouping variables, so that corresponding values in data variables are ignored.

Grouping Variables

Grouping variablesare variables used to group, or categorize, observations—that is, data values in other variables. A grouping variable can be any of these data types:

Numeric, logical, categorical,datetime, ordurationvector
Cell array of character vectors
Table, with table variables of any data type in this list

Data variablesare the variables that contain observations. A grouping variable must have a value corresponding to each value in the data variables. Data values belong to the same group when the corresponding values in the grouping variable are the same.

This table shows examples of data variables, grouping variables, and the groups that you can create when you split the data variables using the grouping variables.

Data Variable	Grouping Variable	Groups of Data
`[5 10 15 20 25 30]`	`[0 0 0 0 1 1]`	`[5 10 15 20] [25 30]`
`[10 20 30 40 50 60]`	`[1 3 3 1 2 1]`	`[10 40 60] [50] [20 30]`
`[64 72 67 69 64 68]`	`{'F','M','F','M','F','F'}`	`[64 67 64 68] [72 69]`

You can give groups of data meaningful names when you use cell arrays of character vectors or categorical arrays as grouping variables. A categorical array is an efficient and flexible choice of grouping variable.

Group Definition

Typically, there are as many groups as there are unique values in the grouping variable. (A categorical array also can include categories that are not represented in the data.) The groups and the order of the groups depend on the data type of the grouping variable.

For numeric, logical,datetime, orduration特征向量的向量,或细胞阵列groups correspond to the unique values sorted in ascending order.
For categorical arrays, the groups correspond to the unique values observed in the array, sorted in the order returned by thecategoriesfunction.

Thefindgroupsfunction can accept multiple grouping variables, for exampleG = findgroups(A1,A2). You also can include multiple grouping variables in a table, for exampleT = table(A1,A2); G = findgroups(T). Thefindgroupsfunction defines groups by the unique combinations of values across corresponding elements of the grouping variables.findgroupsdecides the order by the order of the first grouping variable, and then by the order of the second grouping variable, and so on. For example, ifA1 = {'a','a','b','b'}andA2 = [0 1 0 0], then the unique values across the grouping variables are'a' 0,'a' 1, and'b' 0, defining three groups.

The Split-Apply-Combine Workflow

After you select grouping variables and split data variables into groups, you can apply functions to the groups and combine the results. This workflow is called the Split-Apply-Combine workflow. You can use thefindgroupsandsplitapplyfunctions together to analyze groups of data in this workflow. This diagram shows a simple example using the grouping variable性别and the data variableHeightto calculate the mean height by gender.

Thefindgroupsfunction returns a vector ofgroup numbersthat define groups based on the unique values in the grouping variables.splitapplyuses the group numbers to split the data into groups efficiently before applying a function.

Missing Group Values

Grouping variables can have missing values. This table shows the missing value indicator for each data type. If a grouping variable has missing values, thenfindgroupsassignsNaNas the group number, andsplitapplyignores the corresponding values in the data variables.

Grouping Variable Data Type	Missing Value Indicator
Numeric	`NaN`
Logical	(Cannot be missing)
Categorical
`datetime`	`NaT`
`duration`	`NaN`
Cell array of character vectors	`''`
字符串