Is it possible to have a table with a numerical index instead of strings?

I'm new to MATLAB and wanted to make a table containing time-series data where the index (time) is a series of integers.

I tried this but I think it doesn't like the fact that the row names are not strings.

>> N = 5;
>> k_init = 5;
>> u = zeros(N+k_init,1);
>> y = zeros(N+k_init,1);
>> ts = [1-k_init:N]';
>> data = table(u,y,'RowNames',num2cell(ts));
Error using table (line 326)
The RowNames property must be a string array or a cell array, with each element containing one nonempty name.
>> class(num2cell(ts))

ans =

    'cell'

It seems to only work with strings:

>> ts = {'-4','-3','-2','-1','0','1','2','3','4','5'};
>> data = table(u, y,'RowNames',ts)

data =

  10×2 table

          u    y
          _    _

    -4    0    0
    -3    0    0
    -2    0    0
    -1    0    0
    0     0    0
    1     0    0
    2     0    0
    3     0    0
    4     0    0
    5     0    0
 

What I am ultimately trying to do is access the data using the ts as an index, something like this:

>> k = 2
>> data(k,'y') = 2*data(k-1,'u')

(Similar to a DataFrame in Python)

[Bill Tubbs]

An alternative perhaps: Is there any way to index an iddata object with a timestep value when it doesn't start at 1?

>> data = iddata(y,u,1,'Tstart',-4);
>> data.y(-4)
Error using iddata/subsref (line 49)
Index in position 1 is invalid. Array indices must be positive integers or logical values.

Other than data.u(-4+data.Tstart) obviously.

Answer 1

[Peng Li]

Why do you want to make the time as row names? I think it's better to keep it as one of the variable in the table. It is not necessary that each row has a name.

It's easier to index table by dot indexing. e.g., data.u(k) and data.y(k-1) for example. If you make ts as a variable, you can use for example data.u(data.ts == 2) to index variable u in the table where ts is equal to 2.

[Bill Tubbs]

Ah, interesting. That's one solution. Not quite as concise as data(k,'y') but still good. Thanks.

[Bill Tubbs]

B.t.w. one reason why dataframes (e.g in Python and R) have a dedicated index is because index values have slightly different properties to other column data (e.g. no duplicates).

[Peng Li]

yeah but the index in a matlab table is already there right just implicitely. data.y(1) is the first element of variable y in table data, and so on.

[Bill Tubbs]

I was just making the point that the solution `data.u(data.ts == 2)` could return more than one value if you're not careful.

[Peng Li]

Oh yeah this is correct. So better apply things according to our own understanding of our data.

Answer 2

[Fangjun Jiang]

Keep in mind that in MATLAB, indexing like data(RowIndex, ColIndex) can be numerical or logical. The numerical index is 1-based. It can't be negative numerical value. It can't be zero like in C.

Your value is -4 to 5. There is an easy way to specify an offset.

The other thing you might be able to use is cellstr(num2str((-4:5)'))

[Bill Tubbs]

Thanks, I realise I can use an offset. I just thought there might be a way to do indexing in Matlab when the index doesn't start at 1.

[Bill Tubbs]

Given that this seems to be impossible, I think the best solution might be to give up and just maintain two seperate variables, one for the actual timestep (k=-4:5) and another for the index (i=1:10).

In other words:

>> k = 2
>> i = k - k_start;
>> data(i,'y') = 2*data(i-1,'u')

However, this only works here were the timestep is 1.

Answer 3

[Walter Roberson]

data(string(k),'y') = 2*data(string(k-1),'u')