Calculate mean of all variables that have a mean

I have a table with mixed types of variables. I want to calculate the mean of any variable that can have a mean without converting to another type- so just numericals, logicals, etc. and leave string types alone. I do not want to specify which ones to find the mean for because I will probably add variables to this table eventually. The AI made up this apparently fake option, which is what I am looking for: 'Exclude', {'StringVar'}

Answer 1

[Steven Lord]

Instead of calling varfun twice as the answer posted by @Arun does, call it once with the InputVariables name-value argument. The value of InputVariables can be a function handle that varfun can execute on the table variables; if the function handle returns true varfun will process that variable.

load patients
T = table(LastName,Age,Height,Weight);
T.Over40 = T.Age > 40;
head(T)

      LastName      Age    Height    Weight    Over40
    ____________    ___    ______    ______    ______

    {'Smith'   }    38       71       176      false 
    {'Johnson' }    43       69       163      true  
    {'Williams'}    38       64       131      false 
    {'Jones'   }    40       67       133      false 
    {'Brown'   }    49       64       119      true  
    {'Davis'   }    46       68       142      true  
    {'Miller'  }    33       64       142      false 
    {'Wilson'  }    40       68       180      false 

T2 = varfun(@mean, T, InputVariables=@(x) isnumeric(x) | islogical(x))

T2 = 1x4 table
    mean_Age    mean_Height    mean_Weight    mean_Over40
    ________    ___________    ___________    ___________

     38.28         67.07           154            0.4    

There is no variable mean_LastName in T2 because the LastName variable in T did not satisfy isnumeric | islogical.

[Marcus Glover]

Thanks! This actual intelligence answer is definitely better.

[Paul]

Hi Marcus,

"I want to calculate the mean of any variable that can have a mean"

If that's really the case (and limited in scope to built-in types), then you may need to also check for duration, datetime, and even char if those are potentially relevant.

Interestingly, in 2022a the doc page for mean included a "Data Types" entry that showed which data types are mean-able. For some reason, that information is no longer shown in mean.

[Marcus Glover]

Thanks. Good point- and I do mean it. I often calulate the mode of a char, but never tried the mean... I looked at the documentation and I'm still not even sure what the mean of a set of char would be.

[Paul]

Looks like mean of char means the mean of the array after doing the implicit conversion of char to numeric.

c ='abcd'

c = 'abcd'

[mean(c) mean(double(c))]

ans = 1x2

98.5000 98.5000

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

c =['abcd';'abcd']

c = 2x4 char array
    'abcd'
    'abcd'

mean(c,2)

ans = 2x1

98.5000 98.5000

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

mean(c,1)

ans = 1x4

97 98 99 100

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

double('abcd')

ans = 1x4

97 98 99 100

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

[Marcus Glover]

Thanks! That's probably what mode is doing as well, but it returns the characters

c =['abcd';'abcd';'defg'];
mode(c)

ans = 'abcd'

[Paul]

Interesting. The doc page Mode - Input Argument does not explicitly list char as an acceptable class for the input array. But the doc page Mode - Output Arguments does say that "If the input A is an array, then the output M is an array of the same class."

On the other hand, the doc page for mean is silent on the list of acceptable input arguments, but the discussion under the 'outtype' argument clearly allows for char input. But that section is kind of hard to follow for char input. I guess one could arguy that from that section that char input should result in a double output.

Seems odd that mean and mode would return different class of output for the same input.

Answer 2

[Arun]

Hi @Marcus Glover,

I understand that you wish to calculate the mean of any variable in a table without specifying which ones to target specifically.

To achieve this for variables that naturally support mean calculation (such as numerical and logical types) without needing to explicitly convert types or specify variables, one can programmatically determine the variable types and then calculate the mean for those that are appropriate. This method is flexible and will automatically adjust to new variables added to the table.

Following are the steps that can be followed to achieve the required objective:

1. Step-1: Identify Numerical and Logical Variables: Use the table's “vartype” method or inspect variable types with “varfun” to identify variables that are either numerical or logical, as these are the types for which a mean would typically be meaningful.
2. Step-2: Calculate Mean for Identified Variables:Use the identified variable names to calculate their means.

Here is an example script that demonstrates this approach:

% Example table with mixed variable types
T = table([1; 2; 3; 4], {'a'; 'b'; 'c'; 'd'}, [true; false; true; false], 'VariableNames', {'NumericVar', 'StringVar', 'LogicalVar'});

% Identify variables that are either numerical or logical
isNumericOrLogical = varfun(@(x) isnumeric(x) || islogical(x), T, 'OutputFormat', 'uniform');

% Calculate mean for numerical and logical variables only
means = varfun(@mean, T(:, isNumericOrLogical), 'OutputFormat', 'table');

% Display the result
disp(means);

    mean_NumericVar    mean_LogicalVar
    _______________    _______________

          2.5                0.5      

This method avoids the need for hard-coding variable names or types and will automatically adapt as you add or remove variables from the table, provided one rerun the script or incorporate it into a function that operates on your table.

Please refer the following documentation links for more information regarding the related functions:

1. vartype: https://www.mathworks.com/help/matlab/ref/vartype.html
2. varfun: https://www.mathworks.com/help/matlab/ref/table.varfun.html

Hope this helps!

Regards

Arun

[Marcus Glover]

This AI got it right! Which one was it?

Thanks.