Index and select rows from table

A table includes columns with numerical and categorical columns (see attachement)

How can I get multiple subtables based on the categories of a column (e.g., col5)?

Expected output: 4 different tables (table1includes rows with "h_001" in col5... table4 includes rows with "r_041" in col5)

I could not use the following option when the columns have categorical data:

https://ch.mathworks.com/matlabcentral/answers/481574-how-to-get-the-index-of-a-value-in-a-table

[Dyuman Joshi]

Dynamically naming variables is not recommended - TUTORIAL: Why Variables Should Not Be Named Dynamically (eval)

What do you want to do with the different tables?

[julian gaviria]

Thanks @Dyuman Joshi,

I need to split the original data based on the categories of col5. The following code works:

index1=find(data.Col5=='h_001');
index2=find(data.Col5=='h_022');
index3=find(data.Col5=='h_045');
index4=find(data.Col5=='r_041');
data1=data(index1,:);
data2=data(index2,:);
data3=data(index3,:);
data4=data(index4,:);

However, I wonder whether there is any further alternative to simplify the code. My data has 600 categories in one column of the table.

I am looking for something more efficient than the code above. Perhaps 1 line of code per category?

[Dyuman Joshi]

If you have 600 categories, you will have 600 sub tables. That will be a nightmare to work with.

Did you read the link I attached in my comment above?

I'll reiterate - What do you want to do with the 600 different tables? Are there any operations you would like to perform on them?

[julian gaviria]

@Dyuman Joshi wrote: Are there any operations you would like to perform on them?

Yes, I will obtain a single value from each table...

[Stephen23]

ISMEMBER

[Dyuman Joshi]

"Yes, I will obtain a single value from each table..."

Which value? Max/Min/Median/Mean? or something else?

[julian gaviria]

@Stephen23 I don't see how

ismember()

would be more efficient. It would require at least 2 lines per categorie:

https://ch.mathworks.com/help/matlab/ref/double.ismember.html#d126e855390

[Stephen23]

"I don't see how ismember() would be more efficient. It would require at least 2 lines per categorie"

Whatever way you do it, processing every category individually will be inefficient (in terms of your time writing and/or runtime).

That is why I suggested ISMEMBER, so that you can do all categories at once. Three lines of code, done.

Do not store each category individually. That is not how MATLAB works, you need to learn how to use vectors, matrices, and arrays. Start by placing all of the categories into one array (e.g. a string array), then use one ISMEMBER call. Read the ISMEMBER documentation carefully.

Another option would be to use one of the JOIN family.

[Cris LaPierre]

Do not confuse the number of lines of code with efficiency.

Answer 1

[Peter Perkins]

"I need to split the original data based on the categories of col5."

You probably do not want/need to do that. Take a look at the rowfun function. Write a function to do what you want with each subset of your data, then use rowfun to apply that function based on groups defined by col5.

t = table([1;1;1;2;2;3;3;3],rand(8,1),rand(8,1),VariableNames=["G" "X" "Y"])

t = 8×3 table
    G       X          Y   
    _    _______    _______

    1    0.50908    0.09642
    1    0.46628     0.9575
    1    0.99625    0.40089
    2    0.63605    0.70043
    2    0.81209    0.49475
    3    0.38157    0.51853
    3    0.74877    0.66036
    3    0.85878    0.91136

myFun = @(x,y) mean(x) - mean(y);
rowfun(myFun,t,GroupingVariables="G")

ans = 3×3 table
    G    GroupCount      Var3   
    _    __________    _________

    1        3           0.17227
    2        2           0.12648
    3        3         -0.033704

Lots of other functions, like groupsummary, similarly do not require you to split your data up. As others have said, that's usually a bad idea and unnecessary.