groupsummary

Compute summary statistics by group

Syntax

G = groupsummary(T,groupvars)

G = groupsummary(T,groupvars,method)

G = groupsummary(T,groupvars,method,datavars)

G = groupsummary(T,groupvars,groupbins)

G = groupsummary(T,groupvars,groupbins,method)

G = groupsummary(T,groupvars,groupbins,method,datavars)

G = groupsummary(___,Name,Value)

B = groupsummary(A,groupvars,method)

B = groupsummary(A,groupvars,groupbins,method)

B = groupsummary(___,Name,Value)

[B,BG,BC] = groupsummary(A,___)

Description

Table Data

G = groupsummary(T,groupvars) returns the unique grouping variable combinations and the number of members in each group for table or timetable T. Groups are defined by rows in the variables in groupvars that have the same unique combination of values. Each row of the output table corresponds to one group. For example, G = groupsummary(T,"HealthStatus") returns a table with the count of each group in the variable HealthStatus.

You can use groupsummary functionality interactively by adding the Compute by Group task to a live script.

For more information, see Group Summary Computation.

G = groupsummary(T,groupvars,method) aggregates data using the group-wise computations specified in method and appends the computation results to the output table as additional variables. For example, G = groupsummary(T,"Location","median") returns the median value of every nongrouping variable in T for each location, in addition to the number of members in each location group.

G = groupsummary(T,groupvars,method,datavars) specifies the table variables to apply the method to.

G = groupsummary(T,groupvars,groupbins) specifies to bin rows in groupvars according to binning scheme groupbins prior to grouping. For example, G = groupsummary(T,"SaleDate","year") returns the group counts for all sales in T within each year according to the grouping variable SaleDate.

G = groupsummary(T,groupvars,groupbins,method) specifies how to bin the data in groupvars prior to grouping and applying the group-wise computations specified in method.

G = groupsummary(T,groupvars,groupbins,method,datavars) specifies the table variables to apply the method to.

G = groupsummary(___,Name,Value) specifies additional grouping properties using one or more name-value arguments for any of the previous syntaxes. For example, G = groupsummary(T,"Category1","IncludeMissingGroups",false) excludes the group made from missing data of type categorical indicated by <undefined> in Category1.

Array Data

B = groupsummary(A,groupvars,method) returns the concatenated results of applying the group-wise computations in method to unique groups in vector, matrix, or cell array A. Groups are defined by rows in the column vectors in groupvars that have the same unique combination of values. Each row of the output array contains the computation results for one group.

You can use groupsummary functionality interactively by adding the Compute by Group task to a live script.

B = groupsummary(A,groupvars,groupbins,method) specifies to bin rows in groupvars according to binning scheme groupbins prior to grouping.

B = groupsummary(___,Name,Value) specifies additional grouping properties using one or more name-value arguments for either of the previous syntaxes for an input array.

[B,BG,BC] = groupsummary(A,___) returns additional group information. BG is the unique grouping vector combinations corresponding to the rows in B. BC is the number of members in each group in BG.

Examples

Summary Statistics

Compute summary statistics on table variables.

Create a table T that contains information about eight individuals.

HealthStatus = categorical(["Poor"; "Good"; "Fair"; "Fair"; "Poor"; "Excellent"; "Good"; "Excellent"]);
Age = [38; 43; 38; 40; 49; 51; 52; 35];
Height = [71; 68; 64; 67; 64; 62; 65; 55];
Weight = [176; 153; 131; 133; 119; 120; 140; 129];
T = table(HealthStatus,Age,Height,Weight)

T=8×4 table
    HealthStatus    Age    Height    Weight
    ____________    ___    ______    ______

     Poor           38       71       176  
     Good           43       68       153  
     Fair           38       64       131  
     Fair           40       67       133  
     Poor           49       64       119  
     Excellent      51       62       120  
     Good           52       65       140  
     Excellent      35       55       129  

Compute the counts of the health status groups by specifying HealthStatus as the grouping variable.

G = groupsummary(T,"HealthStatus")

G=4×2 table
    HealthStatus    GroupCount
    ____________    __________

     Excellent          2     
     Fair               2     
     Good               2     
     Poor               2     

Compute the mean age, height, and weight of each health status group.

G = groupsummary(T,"HealthStatus","mean")

G=4×5 table
    HealthStatus    GroupCount    mean_Age    mean_Height    mean_Weight
    ____________    __________    ________    ___________    ___________

     Excellent          2             43         58.5           124.5   
     Fair               2             39         65.5             132   
     Good               2           47.5         66.5           146.5   
     Poor               2           43.5         67.5           147.5   

Still grouping by health status, compute only the median height.

G = groupsummary(T,"HealthStatus","median","Height")

G=4×3 table
    HealthStatus    GroupCount    median_Height
    ____________    __________    _____________

     Excellent          2             58.5     
     Fair               2             65.5     
     Good               2             66.5     
     Poor               2             67.5     

Multiple Grouping Variables

Group table data using two grouping variables.

Create a table T that contains information about eight individuals.

HealthStatus = categorical(["Poor"; "Good"; "Fair"; "Fair"; "Poor"; "Excellent"; "Good"; "Excellent"]);
Smoker = logical([1; 0; 0; 1; 1; 0; 0; 1]);
Weight = [176; 153; 131; 133; 119; 120; 140; 129];
T = table(HealthStatus,Smoker,Weight)

T=8×3 table
    HealthStatus    Smoker    Weight
    ____________    ______    ______

     Poor           true       176  
     Good           false      153  
     Fair           false      131  
     Fair           true       133  
     Poor           true       119  
     Excellent      false      120  
     Good           false      140  
     Excellent      true       129  

Compute the mean weight, grouped by health status and smoking status. By default, some combinations of health status and smoking status are not represented in the output because they are empty groups.

G = groupsummary(T,["HealthStatus","Smoker"],"mean","Weight")

G=6×4 table
    HealthStatus    Smoker    GroupCount    mean_Weight
    ____________    ______    __________    ___________

     Excellent      false         1              120   
     Excellent      true          1              129   
     Fair           false         1              131   
     Fair           true          1              133   
     Good           false         2            146.5   
     Poor           true          2            147.5   

Set the value of IncludeEmptyGroups to true to see all group combinations, including the empty ones.

G = groupsummary(T,["HealthStatus","Smoker"],"mean","Weight","IncludeEmptyGroups",true)

G=8×4 table
    HealthStatus    Smoker    GroupCount    mean_Weight
    ____________    ______    __________    ___________

     Excellent      false         1              120   
     Excellent      true          1              129   
     Fair           false         1              131   
     Fair           true          1              133   
     Good           false         2            146.5   
     Good           true          0              NaN   
     Poor           false         0              NaN   
     Poor           true          2            147.5   

Specify Group Bins

Group data according to specified bins.

Create a timetable that contains sales information for days within a single month.

TimeStamps = datetime([2017 3 4; 2017 3 2; 2017 3 15; 2017 3 10; ...
                       2017 3 14; 2017 3 31; 2017 3 25; ...
                       2017 3 29; 2017 3 21; 2017 3 18]);
Profit = [2032 3071 1185 2587 1998 2899 3112 909 2619 3085]';
ItemsSold = [14 13 8 5 10 16 8 6 7 11]';
TT = timetable(TimeStamps,Profit,ItemsSold)

TT=10×2 timetable
    04-Mar-2017    2032    14
    02-Mar-2017    3071    13
    15-Mar-2017    1185    8
    10-Mar-2017    2587    5
    14-Mar-2017    1998    10
    31-Mar-2017    2899    16
    25-Mar-2017    3112    8
    29-Mar-2017    909    6
    21-Mar-2017    2619    7
    18-Mar-2017    3085    11

Compute the mean and the mode of profit binned by the items sold, binning the groups into intervals of item numbers.

format shorte
G = groupsummary(TT,"ItemsSold",[0 4 8 12 16],{"mean","mode"},"Profit")

G=3×4 table
    [4, 8)    3    2.0383e+03    909
    [8, 12)    4    2.3450e+03    1185
    [12, 16]    3    2.6673e+03    2032

Compute the mean profit grouped by day of the week.

G = groupsummary(TT,"TimeStamps","dayname","mean","Profit")

G=5×3 table
    Tuesday    2    2.3085e+03
    Wednesday    2    1.0470e+03
    Thursday    1    3.0710e+03
    Friday    2    2.7430e+03
    Saturday    3    2.7430e+03

Group Operations with Vector Data

Create a vector of dates and a vector of corresponding profit values.

TimeStamps = datetime([2017 3 4; 2017 3 2; 2017 3 15; 2017 3 10; ...
                       2017 3 14; 2017 3 31; 2017 3 25; ...
                       2017 3 29; 2017 3 21; 2017 3 18]);
Profit = [2032 3071 1185 2587 1998 2899 3112 909 2619 3085]';

Compute the mean profit for each day of the week. Display the mean, the group names, and the number of members in each group.

format shorte
[meanDailyProfit,dayOfWeek,dailyCounts] = groupsummary(Profit,TimeStamps,"dayname","mean")

meanDailyProfit = 5×1

   2.3085e+03
   1.0470e+03
   3.0710e+03
   2.7430e+03
   2.7430e+03

dayOfWeek = 5×1 categorical
     Tuesday 
     Wednesday 
     Thursday 
     Friday 
     Saturday 

dailyCounts = 5×1

     2
     2
     1
     2
     3

Multiple Grouping Vectors for Vector Input

Compute the mean weights for a set of people grouped by their health status and smoker status.

Store information about the individuals as three vectors of different types.

HealthStatus = categorical(["Poor"; "Good"; "Fair"; "Fair"; "Poor"; "Excellent"; "Good"; "Excellent"]);
Smoker = logical([1; 0; 0; 1; 1; 0; 0; 1]);
Weight = [176; 153; 131; 133; 119; 120; 140; 129];

Grouping by health status and smoker status, compute the mean weights.

B contains the mean for each group (NaN for empty groups). BG is a cell array containing two vectors that describe the groups as you look at their elements row-wise. For instance, the first row of BG{1} indicates that the patients in the first group have a health status Excellent, and the first row of BG{2} indicates that they are nonsmokers. Finally, BC contains the number of members in each group for the corresponding groups in BG.

[B,BG,BC] = groupsummary(Weight,{HealthStatus,Smoker},"mean","IncludeEmptyGroups",true);
B

B = 8×1

  120.0000
  129.0000
  131.0000
  133.0000
  146.5000
       NaN
       NaN
  147.5000

BG{1}

ans = 8×1 categorical
     Excellent 
     Excellent 
     Fair 
     Fair 
     Good 
     Good 
     Poor 
     Poor 

BG{2}

ans = 8×1 logical array

   0
   1
   0
   1
   0
   1
   0
   1

BC

BC = 8×1

     1
     1
     1
     1
     2
     0
     0
     2

Method Function Handle with Multiple Inputs

Load data containing patient information and create a table describing each patient's location, systolic and diastolic blood pressure, height, and weight.

load patients
Location = categorical(Location);
T = table(Location,Systolic,Diastolic,Height,Weight)

T=100×5 table
            Location             Systolic    Diastolic    Height    Weight
    _________________________    ________    _________    ______    ______

    County General Hospital        124          93          71       176  
    VA Hospital                    109          77          69       163  
    St. Mary's Medical Center      125          83          64       131  
    VA Hospital                    117          75          67       133  
    County General Hospital        122          80          64       119  
    St. Mary's Medical Center      121          70          68       142  
    VA Hospital                    130          88          64       142  
    VA Hospital                    115          82          68       180  
    St. Mary's Medical Center      115          78          68       183  
    County General Hospital        118          86          66       132  
    County General Hospital        114          77          68       128  
    St. Mary's Medical Center      115          68          66       137  
    VA Hospital                    127          74          71       174  
    VA Hospital                    130          95          72       202  
    St. Mary's Medical Center      114          79          65       129  
    VA Hospital                    130          92          71       181  
      ⋮

Grouping by location, compute the correlation between patient height and weight and the correlation between systolic and diastolic blood pressure. Use the xcov function as the method to compute the correlation. The first two input arguments to xcov describe the data to correlate, the third argument describes the lag size, and the fourth argument describes the type of normalization. For each group computation, the x and y arguments passed to xcov are specified pairwise by variable from the two cell elements ["Height","Systolic"] and ["Weight","Diastolic"].

G = groupsummary(T,"Location",@(x,y) xcov(x,y,0,"coeff"),{["Height","Systolic"],["Weight","Diastolic"]})

G=3×4 table
            Location             GroupCount    fun1_Height_Weight    fun1_Systolic_Diastolic
    _________________________    __________    __________________    _______________________

    County General Hospital          39             0.65483                  0.44187        
    St. Mary's Medical Center        24             0.62047                  0.44466        
    VA Hospital                      37             0.78438                  0.62256        

Alternatively, if your data is in vector or matrix form instead of in a table, you can provide the data to correlate as the first input argument of groupsummary.

[B,BG,BC] = groupsummary({[Height,Systolic],[Weight,Diastolic]},Location,@(x,y) xcov(x,y,0,"coeff"))

B = 3×2

    0.6548    0.4419
    0.6205    0.4447
    0.7844    0.6226

BG = 3×1 categorical
     County General Hospital 
     St. Mary's Medical Center 
     VA Hospital 

BC = 3×1

    39
    24
    37

Input Arguments

T — Input table

table | timetable

Input table, specified as a table or timetable.

A — Input array

column vector | matrix | cell array

Input array, specified as a column vector, group of column vectors stored as a matrix, or cell array of column vectors, character row vectors, or matrices.

When you specify a function handle for method that takes more than one input argument, the input array A must be a cell array of column vectors, character row vectors, or matrices. In each call to the function by group, the input arguments are the corresponding columns of each element in the cell array. For example:

• groupsummary({x1,y1},groupvars,@(x,y) myFun(x,y)) calculates myFun(x1,y1) for each group.

• groupsummary({[x1 x2],[y1 y2]},groupvars,@(x,y) myFun(x,y)) first calculates myFun(x1,y1) for each group and then calculates myFun(x2,y2) for each group.

groupvars — Grouping variables or vectors

scalar | vector | matrix | cell array | pattern | function handle | table vartype subscript

Grouping variables or vectors, specified as one of these options:

• For array input data, groupvars can be either a column vector with the same number of rows as A or a group of column vectors arranged in a matrix or cell array.

• For table or timetable input data, groupvars indicates which variables to use to compute groups in the data. You can specify the grouping variables with any of the options in this table.

  Indexing Scheme	Values to Specify	Examples
  Variable name	A string scalar or character vector A string array or cell array of character vectors A pattern object	"A" or 'A' — A variable named A ["A" "B"] or {'A','B'} — Two variables named A and B "Var"+digitsPattern(1) — Variables named "Var" followed by a single digit
  Variable index	An index number that refers to the location of a variable in the table A vector of numbers A logical vector. Typically, this vector is the same length as the number of variables, but you can omit trailing 0 (false) values.	3 — The third variable from the table [2 3] — The second and third variables from the table [false false true] — The third variable
  Function handle	A function handle that takes a table variable as input and returns a logical scalar	@isnumeric — All the variables containing numeric values
  Variable type	A vartype subscript that selects variables of a specified type	vartype("numeric") — All the variables containing numeric values

Example: groupsummary(T,"Var3")

method — Computation method

"sum" | "mean" | "median" | "all" | function handle | cell array | ...

Computation method, specified as one of these values:

Method	Description
"sum"	Sum
"mean"	Mean
"median"	Median
"mode"	Mode
"var"	Variance
"std"	Standard deviation
"min"	Minimum
"max"	Maximum
"range"	Maximum minus minimum
"nummissing"	Number of missing elements
"numunique"	Number of distinct nonmissing elements
"nnz"	Number of nonzero and non-NaN elements
"all"	All computations previously listed

You also can specify method as a function handle that returns one output per group whose first dimension has length 1. For table input data, the function operates on each table variable separately.

When the input data is a table T and you specify a function handle for method that takes more than one input argument, you must specify datavars. The datavars argument must be a cell array whose elements indicate the table variables to use for each input into the method. In each call to the function by group, the input arguments are the corresponding table variables of the cell array elements. For example:

• groupsummary(T,groupvars,@(x,y) myFun(x,y),{"x1","y1"}) calculates myFun(T.x1,T.y1) for each group.

• groupsummary(T,groupvars,@(x,y) myFun(x,y),{["x1" "x2"],["y1" "y2"]}) first calculates myFun(T.x1,T.y1) for each group and then calculates myFun(T.x2,T.y2) for each group.

When the input data is a vector or matrix and you specify a function handle for method that takes more than one input argument, the input data A must be a cell array of vectors or matrices. In each call to the function, the input arguments are the corresponding columns of each element in the cell array. For example:

• groupsummary({x1,y1},groupvars,@(x,y) myFun(x,y)) calculates myFun(x1,y1) for each group.

• groupsummary({[x1 x2],[y1 y2]},groupvars,@(x,y) myFun(x,y)) first calculates myFun(x1,y1) for each group and then calculates myFun(x2,y2) for each group.

To specify multiple computations at a time, list the options in a cell array, such as {"mean","median"} or {myFun1,myFun2}.

NaN values in the input data are automatically omitted when using the method names described here, with the exception of "nummissing". To include NaN values, use a function handle for the method, such as @sum instead of "sum".

Data Types: char | string | cell | function_handle

datavars — Table variables to operate on

scalar | vector | cell array | function handle | pattern | table vartype subscript

Table variables to operate on, specified as one of the options in this table. datavars indicates which variables of the input table or timetable to apply the methods to. Other variables not specified by datavars are not operated on and do not pass through to the output. When datavars is not specified, groupsummary operates on each nongrouping variable.

Indexing Scheme	Values to Specify	Examples
Variable name	A string scalar or character vector A string array or cell array of character vectors A pattern object	"A" or 'A' — A variable named A ["A" "B"] or {'A','B'} — Two variables named A and B "Var"+digitsPattern(1) — Variables named "Var" followed by a single digit
Variable index	An index number that refers to the location of a variable in the table A vector of numbers A logical vector. Typically, this vector is the same length as the number of variables, but you can omit trailing 0 (false) values.	3 — The third variable from the table [2 3] — The second and third variables from the table [false false true] — The third variable
Function handle	A function handle that takes a table variable as input and returns a logical scalar	@isnumeric — All the variables containing numeric values
Variable type	A vartype subscript that selects variables of a specified type	vartype("numeric") — All the variables containing numeric values

When the input data is a table T and you specify a function handle for method that takes more than one input argument, you must specify datavars. The datavars argument must be a cell array whose elements are any of the options in the table. The cell array elements indicate the table variables to use for each input into the method. In each call to the function by group, the input arguments are the corresponding table variables of the cell array elements. For example:

• groupsummary(T,groupvars,@(x,y) myFun(x,y),{"x1","y1"}) calculates myFun(T.x1,T.y1) for each group.

• groupsummary(T,groupvars,@(x,y) myFun(x,y),{["x1" "x2"],["y1" "y2"]}) first calculates myFun(T.x1,T.y1) for each group and then calculates myFun(T.x2,T.y2) for each group.

Example: groupsummary(T,groupvars,method,["Var1" "Var2" "Var4"])

groupbins — Binning scheme for grouping variables or vectors

"none" (default) | vector of bin edges | number of bins | length of time (bin width) | name of time unit (bin width) | cell array of binning methods

Binning scheme for grouping variables or vectors, specified as one or more of the following binning methods. Grouping variables or vectors and binning scheme arguments must be the same size, or one of them can be scalar.

• "none" — No binning.

• Vector of bin edges — The bin edges define the bins. You can specify the edges as numeric values or as datetime values for datetime grouping variables or vectors.

• Number of bins — The number determines how many equally spaced bins to create. You can specify the number of bins as a positive integer scalar.

• Length of time (bin width) — The length of time determines the width of each bin. You can specify the bin width as a duration or calendarDuration scalar for datetime or duration grouping variables or vectors.

• Name of time unit (bin width) — The name of the time unit determines the width of each bin. You can specify the bin width as one of the options in this table for datetime or duration grouping variables or vectors.

  Value	Description	Data Type
  "second"	Each bin is 1 second.	datetime and duration
  "minute"	Each bin is 1 minute.	datetime and duration
  "hour"	Each bin is 1 hour.	datetime and duration
  "day"	Each bin is 1 calendar day. This value accounts for daylight saving time shifts.	datetime and duration
  "week"	Each bin is 1 calendar week.	datetime only
  "month"	Each bin is 1 calendar month.	datetime only
  "quarter"	Each bin is 1 calendar quarter.	datetime only
  "year"	Each bin is 1 calendar year. This value accounts for leap days.	datetime and duration
  "decade"	Each bin is 1 decade (10 calendar years).	datetime only
  "century"	Each bin is 1 century (100 calendar years).	datetime only
  "secondofminute"	Bins are seconds from 0 to 59.	datetime only
  "minuteofhour"	Bins are minutes from 0 to 59.	datetime only
  "hourofday"	Bins are hours from 0 to 23.	datetime only
  "dayofweek"	Bins are days from 1 to 7. The first day of the week is Sunday.	datetime only
  "dayname"	Bins are full day names, such as "Sunday".	datetime only
  "dayofmonth"	Bins are days from 1 to 31.	datetime only
  "dayofyear"	Bins are days from 1 to 366.	datetime only
  "weekofmonth"	Bins are weeks from 1 to 6.	datetime only
  "weekofyear"	Bins are weeks from 1 to 54.	datetime only
  "monthname"	Bins are full month names, such as "January".	datetime only
  "monthofyear"	Bins are months from 1 to 12.	datetime only
  "quarterofyear"	Bins are quarters from 1 to 4.	datetime only

Example: G = groupsummary(T,"Var1",[-Inf 0 Inf])

Example: G = groupsummary(T,["Var1" "Var2"],{"none" "year"})

Name-Value Arguments

Specify optional pairs of arguments as Name1=Value1,...,NameN=ValueN, where Name is the argument name and Value is the corresponding value. Name-value arguments must appear after other arguments, but the order of the pairs does not matter.

Example: G = groupsummary(T,groupvars,groupbins,IncludedEdge="right")

Before R2021a, use commas to separate each name and value, and enclose Name in quotes.

Example: G = groupsummary(T,groupvars,groupbins,"IncludedEdge","right")

IncludedEdge — Included bin edge for binning scheme

"left" (default) | "right"

Included bin edge for binning scheme, specified as either "left" or "right", indicating which end of the bin interval is inclusive.

You can specify IncludedEdge only if you also specify groupbins, and the value applies to all binning methods for all grouping variables or vectors.

IncludeMissingGroups — Option to treat missing values as a group

true or 1 (default) | false or 0

Option to treat missing values as a group, specified as a numeric or logical 1 (true) or 0 (false). If IncludeMissingGroups is true, then groupsummary treats missing values, such as NaN, in a grouping variable or vector as a group. If a grouping variable or vector has no missing values, or if IncludeMissingGroups is false, then groupsummary does not treat missing values as a group.

IncludeEmptyGroups — Option to include empty groups in group summary operation

false or 0 (default) | true or 1

Option to include empty groups in the group summary operation, specified as a numeric or logical 0 (false) or 1 (true). If IncludeEmptyGroups is false, then groupsummary omits empty groups. If IncludeEmptyGroups is true, then groupsummary includes empty groups.

An empty group occurs in these cases:

• A possible value of a grouping variable or vector is not represented in the input data, such as in a categorical, logical, or binned numeric variable or vector. For example, if no row in the input table has a value of true for a logical grouping variable, then true defines an empty group.

• A unique combination of grouping variables or vectors is not represented in the input data. For example, if there is no row in the input table where the value of grouping variable A is A1 and the value of grouping variable B is B1, then A1_B1 defines an empty group.

Output Arguments

G — Output table

table

Output table for table or timetable input data, returned as a table. G contains the computed groups, the number of elements in each group, and if method is provided, the result of the specified computations.

B — Output array

vector | matrix

Output array for array input data, returned as a vector or matrix. B contains the specified computations for each group. When multiple methods are specified, groupsummary horizontally concatenates the computations in the order that they were listed.

BG — Groups

column vector | cell array of column vectors

Groups for array input data, returned as a column vector or cell array of column vectors. For a single grouping vector, the output groups are sorted according to the order returned by the unique function with the "sorted" option.

For more than one input vector, BG is a cell array containing column vectors of equal length. Information for each group is contained in the elements of a row across all vectors in BG. Each group maps to the corresponding row of the output array B.

BC — Group counts

column vector

Group counts for array input data, returned as a column vector. BC contains the number of elements in each group. The length of BC is the same as the length of the group column vectors returned in BG.

More About

Group Summary Computation

This table illustrates group summary computations.

Sample Table T	Syntax Example	Resulting Table
	groupsummary(T,"VarA")
	groupsummary(T,"VarA","mean")
	groupsummary(T,["VarA" "VarB"],{"none",[-Inf 0 Inf]},"min")
	groupsummary(T,"VarA",["mean" "median" "mode"],"VarB")

Tips

• When making many calls to groupsummary, consider converting grouping variables to type categorical or logical when possible for improved performance. For example, if you have a string array grouping variable (such as HealthStatus with elements "Poor", "Fair", "Good", and "Excellent"), you can convert it to a categorical variable using the command categorical(HealthStatus).

• The groupsummary function computes one-dimensional summary statistics. To compute grouped summaries in two dimensions, consider using the pivot function.

Alternative Functionality

Live Editor Task

You can use groupsummary functionality interactively by adding the Compute by Group task to a live script.

Extended Capabilities

Tall Arrays

Calculate with arrays that have more rows than fit in memory.

The groupsummary function supports tall arrays with the following usage notes and limitations:

• If A and groupvars are both tall matrices, then they must have the same number of rows.

• If the first input is a tall matrix, then groupvars can be a cell array containing tall grouping vectors.

• The groupvars and datavars arguments do not support function handles.

• The IncludeEmptyGroups name-value argument is not supported.

• The "median", "mode", and "numunique" methods are not supported and are not included when specifying the "all" method.

• For tall datetime arrays, the "std" method is not supported.

• If the method argument is a function handle, then it must be a valid input for splitapply operating on a tall array. If the function handle takes multiple inputs, then the first input to groupsummary must be a tall table.

• The order of the groups might be different compared to in-memory groupsummary calculations.

• When grouping by discretized datetime arrays, the categorical group names are different compared to in-memory groupsummary calculations.

For more information, see Tall Arrays.

C/C++ Code Generation

Generate C and C++ code using MATLAB® Coder™.

Usage notes and limitations:

• Sparse inputs are not supported.

• Binning scheme is not supported for datetime or duration data.

• Input data that is a heterogeneous cell array with variable-sized contents is not supported.

• Input tables that contain multidimensional arrays are not supported.

• Input data that contains cell arrays of character vectors or cell arrays of cell arrays is not supported.

• Computation methods must be constant.

• Grouping variables must be constant when the first input argument is a table.

• Data variables must be constant.

• Data variables cannot be nested tables.

• Binning scheme specified as character vectors or strings must be constant.

• Name-value arguments must be constant.

• Computation methods cannot return sparse, multidimensional, or cell array results.

• If the number of group variables can change at runtime, the second output BG is a cell array.

GPU Code Generation

Generate CUDA® code for NVIDIA® GPUs using GPU Coder™.

Refer to the usage notes and limitations in the C/C++ Code Generation section. The same usage notes and limitations apply to GPU code generation.

Thread-Based Environment

Run code in the background using MATLAB® backgroundPool or accelerate code with Parallel Computing Toolbox™ ThreadPool.

The groupsummary function fully supports thread-based environments. For more information, see Run MATLAB Functions in Thread-Based Environment.

Version History

Introduced in R2018a

R2026a: Improved performance for numeric or string grouping vector

The groupsummary function shows improved performance when the grouping variable or vector type is numeric or string.

When the grouping variable or vector is numeric, the improvement is most noticeable when the total number of elements and the number of elements per group are large. When the grouping variable or vector type is string, the improvement is most noticeable when the total number of elements is large and the number of elements per group is small.

For example, this code computes the group-wise mean for a numeric vector of length 12,500,000 with 25 groups. The code is about 7.1x faster than in the previous release.

function timingTest
numberOfGroups = 25;
elementsPerGroup = 5e5;
groups = repmat(1:numberOfGroups,1,elementsPerGroup)';
data = randn(numel(groups),1);
G = @() groupsummary(data,groups,"mean");
t = timeit(G)
end

The approximate execution times are:

R2025b: 1.35 s

R2026a: 0.19 s

The code was timed on a Windows^® 11, AMD EPYC™ 74F3 24-Core Processor @ 3.19 GHz test system by calling the timingTest function.

R2025a: Improved performance when applying multiple computation methods

The groupsummary function shows improved performance when you specify at least two of the "sum", "mean", "min", or "max" computation methods, especially when the number of data points per group is small.

For example, this code computes the sum and maximum for a table with one data variable and one grouping variable. The code is about 6.6x faster than in the previous release.

function timingTest
data = rand(1000,1);
groups = categorical(randi(500,[1000 1]));
for i = 1:1200
   groupsummary(data,groups,["sum" "max"]);
end
end

The approximate execution times are:

R2024b: 2.11 s

R2025a: 0.32 s

The code was timed on a Windows 11, AMD EPYC 74F3 24-Core Processor @ 3.19 GHz test system using the timeit function.

timeit(@timingTest)

R2024a: Apply multiple binning methods to grouping variable

Apply multiple binning methods to one grouping variable or vector by specifying a cell array of binning methods.

R2023a: Return the number of unique elements

Compute the number of distinct nonmissing elements in each group of data. Specify the "numunique" or "all" computation method.

The "all" computation method now returns the number of unique values in addition the computation methods in the previous release.

R2022b: Character arrays have no standard missing value

Character arrays have no default definition of a standard missing value. Therefore, the nummissing method treats blank character array elements (' ') as nonmissing.

R2022a: Code generation support

Generate C or C++ code for the groupsummary function. For usage notes and limitations, see C/C++ Code Generation.

R2022a: Improved performance with small group size

The groupsummary function shows improved performance, especially when the data count in each group is small.

For example, this code performs group summary computations on a matrix with 500 groups with a count of 10 each. The code is about 2.70x faster than in the previous release.

function timingGroupsummary
data = (1:5000)';
groups = repelem(1:length(data)/10,10)';
p = randperm(length(data));
data = data(p);
groups = groups(p);

tic
for k = 1:300
   G = groupsummary(data,groups,"mean");
end
toc
end

The approximate execution times are:

R2021b: 2.65 s

R2022a: 0.98 s

The code was timed on a Windows 10, Intel^® Xeon^® CPU E5-1650 v4 @ 3.60 GHz test system by calling the timingGroupsummary function.

R2022a: Accept data types with no standard missing value

The "nummissing" and "nnz" methods no longer error for input data types with no default definition of a standard missing value.

Code that relied on the errors that MATLAB threw for those inputs, such as code within a try/catch block, may no longer catch those errors.