You can use the table data type to collect mixed-type data and metadata properties, such as variable name, row names, descriptions, and variable units, in a single container. Tables are suitable for column-oriented or tabular data that is often stored as columns in a text file or in a spreadsheet. For example, you can use a table to store experimental data, with rows representing different observations and columns representing different measured variables.
Tables consist of rows and column-oriented variables. Each variable in a table can have a different data type and a different size, but each variable must have the same number of rows.
For example, load sample patients data.
load patientsThen, combine the workspace variables, Systolic and Diastolic into a single BloodPressure variable and convert the workspace variable, Gender, from a cell array of character vectors to a categorical array.
BloodPressure = [Systolic Diastolic]; Gender = categorical(Gender); whos('Gender','Age','Smoker','BloodPressure')
Name Size Bytes Class Attributes Age 100x1 800 double BloodPressure 100x2 1600 double Gender 100x1 330 categorical Smoker 100x1 100 logical
The variables Age, BloodPressure, Gender, and Smoker have varying data types and are candidates to store in a table since they all have the same number of rows, 100.
Now, create a table from the variables and display the first five rows.
T = table(Gender,Age,Smoker,BloodPressure); T(1:5,:)
ans=5×4 table
Gender Age Smoker BloodPressure
______ ___ ______ _____________
Male 38 true 124 93
Male 43 false 109 77
Female 38 false 125 83
Female 40 false 117 75
Female 49 false 122 80
The table displays in a tabular format with the variable names at the top.
Each variable in a table is a single data type. If you add a new row to the table, MATLAB® forces consistency of the data type between the new data and the corresponding table variables. For example, if you try to add information for a new patient where the first column contains the patient's age instead of gender, as in the expression T(end+1,:) = {37,{'Female'},true,[130 84]}, then you receive the error:
Invalid RHS for assignment to a categorical array.
The error occurs because MATLAB® cannot assign numeric data, 37, to the categorical array, Gender.
For comparison of tables with structures, consider the structure array, StructArray, that is equivalent to the table, T.
StructArray = table2struct(T)
StructArray=100×1 struct array with fields:
Gender
Age
Smoker
BloodPressure
Structure arrays organize records using named fields. Each field's value can have a different data type or size. Now, display the named fields for the first element of StructArray.
StructArray(1)
ans = struct with fields:
Gender: Male
Age: 38
Smoker: 1
BloodPressure: [124 93]
Fields in a structure array are analogous to variables in a table. However, unlike with tables, you cannot enforce homogeneity within a field. For example, you can have some values of S.Gender that are categorical array elements, Male or Female, others that are character vectors, 'Male' or 'Female', and others that are integers, 0 or 1.
Now consider the same data stored in a scalar structure, with four fields each containing one variable from the table.
ScalarStruct = struct(... 'Gender',{Gender},... 'Age',Age,... 'Smoker',Smoker,... 'BloodPressure',BloodPressure)
ScalarStruct = struct with fields:
Gender: [100x1 categorical]
Age: [100x1 double]
Smoker: [100x1 logical]
BloodPressure: [100x2 double]
Unlike with tables, you cannot enforce that the data is rectangular. For example, the field ScalarStruct.Age can be a different length than the other fields.
A table allows you to maintain the rectangular structure (like a structure array) and enforce homogeneity of variables (like fields in a scalar structure). Although cell arrays do not have named fields, they have many of the same disadvantages as structure arrays and scalar structures. If you have rectangular data that is homogeneous in each variable, consider using a table. Then you can use numeric or named indexing, and you can use table properties to store metadata.
You can index into a table using parentheses, curly braces, or dot indexing. Parentheses allow you to select a subset of the data in a table and preserve the table container. Curly braces and dot indexing allow you to extract data from a table. Within each table indexing method, you can specify the rows or variables to access by name or by numeric index.
Consider the sample table from above. Each row in the table, T, represents a different patient. The workspace variable, LastName, contains unique identifiers for the 100 rows. Add row names to the table by setting the RowNames property to LastName and display the first five rows of the updated table.
T.Properties.RowNames = LastName; T(1:5,:)
ans=5×4 table
Gender Age Smoker BloodPressure
______ ___ ______ _____________
Smith Male 38 true 124 93
Johnson Male 43 false 109 77
Williams Female 38 false 125 83
Jones Female 40 false 117 75
Brown Female 49 false 122 80
In addition to labeling the data, you can use row and variable names to access data in the table. For example, use named indexing to display the age and blood pressure of the patients Williams and Brown.
T({'Williams','Brown'},{'Age','BloodPressure'})ans=2×2 table
Age BloodPressure
___ _____________
Williams 38 125 83
Brown 49 122 80
Now, use numeric indexing to return an equivalent subtable. Return the third and fifth row from the second and fourth variables.
T(3:2:5,2:2:4)
ans=2×2 table
Age BloodPressure
___ _____________
Williams 38 125 83
Brown 49 122 80
With cell arrays or structures, you do not have the same flexibility to use named or numeric indexing.
With a cell array, you must use strcmp to find desired named data, and then you can index into the array.
With a scalar structure or structure array, it is not possible to refer to a field by number. Furthermore, with a scalar structure, you cannot easily select a subset of variables or a subset of observations. With a structure array, you can select a subset of observations, but you cannot select a subset of variables.
With a table, you can access data by named index or by numeric index. Furthermore, you can easily select a subset of variables and a subset of rows.
For more information on table indexing, see Access Data in Tables.
In addition to storing data, tables have properties to store metadata, such as variable names, row names, descriptions, and variable units. You can access a property using T.Properties.PropName, where T is the name of the table and PropName is one of the table properties.
For example, add a table description, variable descriptions, and variable units for Age.
T.Properties.Description = 'Simulated Patient Data'; T.Properties.VariableDescriptions = ... {'Male or Female' ... '' ... 'true or false' ... 'Systolic/Diastolic'}; T.Properties.VariableUnits{'Age'} = 'Yrs';
Individual empty character vectors within the cell array for VariableDescriptions indicate that the corresponding variable does not have a description. For more information, see the Properties section of table.
To print a table summary, use the summary function.
summary(T)
Description: Simulated Patient Data
Variables:
Gender: 100x1 categorical
Properties:
Description: Male or Female
Values:
Female 53
Male 47
Age: 100x1 double
Properties:
Units: Yrs
Values:
Min 25
Median 39
Max 50
Smoker: 100x1 logical
Properties:
Description: true or false
Values:
True 34
False 66
BloodPressure: 100x2 double
Properties:
Description: Systolic/Diastolic
Values:
Column 1 Column 2
________ ________
Min 109 68
Median 122 81.5
Max 138 99
Structures and cell arrays do not have properties for storing metadata.