Problem with direct calculation on table with std and "omitnan"

Since R2023a, it is possible to perform calculations directly on tables (and timetables) without extracting their data by indexing.

https://fr.mathworks.com/help/matlab/matlab_prog/direct-calculations-on-tables-and-timetables.html?searchHighlight=table&s_tid=srchtitle_table_9

I want use std directly on a numeric table where I can have nan.

For example :

load patients

T = table(Age,Height,Weight,Systolic,Diastolic)

mean(T,"omitnan")

It's fine.

But why there is a problem with std(T,"omitnan") ?

% Applying the function 'std' to the variable 'Age' generated an error.

I can use std(T{:,:},"omitnan") or std(T.Variables,"omitnan") but I lost the possibility to work directly with my table.

Did I miss something ?

Do you have any suggestion ?

Thank you in advance.

SAINTHILLIER Jean Marie

Answer 1

[Star Strider]

Prehaps varfun?

Example —

load patients 
T = table(Age,Height,Weight,Systolic,Diastolic)

T = 100×5 table
    Age    Height    Weight    Systolic    Diastolic
    ___    ______    ______    ________    _________

    38       71       176        124          93    
    43       69       163        109          77    
    38       64       131        125          83    
    40       67       133        117          75    
    49       64       119        122          80    
    46       68       142        121          70    
    33       64       142        130          88    
    40       68       180        115          82    
    28       68       183        115          78    
    31       66       132        118          86    
    45       68       128        114          77    
    42       66       137        115          68    
    25       71       174        127          74    
    39       72       202        130          95    
    36       65       129        114          79    
    48       71       181        130          92    

S = varfun(@(x)std(x,'omitnan'),T)

S = 1×5 table
    Fun_Age    Fun_Height    Fun_Weight    Fun_Systolic    Fun_Diastolic
    _______    __________    __________    ____________    _____________

    7.2154       2.8365        26.571         6.7128          6.9325    

.

[Jean-Marie Sainthillier]

It's strange, it works with varfun but not direclty.

Thank you Star Strider.

[Star Strider]

As always, my pleasure!

[Keith Goatman]

I too was caught out by the inconsistent behaviour of mean and std on tables in 2024a. The answer above works around it but it doesn't explain why they should behave differently.

Answer 2

[Steven Lord]

This looks like a bug to me. It seems that std for table arrays assumes that if you're passing the missingflag input that you've also specified the normalization option or weights. But specifying just the data and the missingflag input is a supported syntax for a double array input and so probably should be supported for a table array input.

A workaround is to specify the default normalization option (0) in the std call as the second input.

load patients 
T = table(Age,Height,Weight,Systolic,Diastolic);
mean(T,"omitnan")

ans = 1x5 table
     Age     Height    Weight    Systolic    Diastolic
    _____    ______    ______    ________    _________

    38.28    67.07      154       122.78       82.96  

std(T, 0, "omitnan")

ans = 1x5 table
     Age      Height    Weight    Systolic    Diastolic
    ______    ______    ______    ________    _________

    7.2154    2.8365    26.571     6.7128      6.9325  

% check
std(T.Age, "omitnan")

ans = 7.2154

I'll report this to the development staff.