How to convert dataset variables from cell?

I've got a dataset object called data.

data =

subject subid sex age

'rachel' 's1' 'f' 21

'tom' 's2' 'm' 24

-----------------------------------------------------------------------------

QUESTION 1

If I check the class of any of my variables, they're all cell:

class(data.subject), ans=cell

I can't seem to figure out how to convert these from cell to a sensible format for indexing.

For example, if I try:

>> nominal(data.sex)

Error using categorical (line 110) Could not compute unique values in A using UNIQUE.

Error in nominal (line 93) b = b@categorical(varargin{:});

Caused by: Error using cell/unique (line 95) Input A must be a cell array of strings.

And it won't let me convert data.sex to a cell array of strings...

>> cellstr(Testdata.sex) Error using cellstr (line 34) Input must be a string.

>> str(Testdata.sex) Undefined function 'str' for input arguments of type 'cell'.

Any suggestions?

Is this part of the problem?

>> data.subject

    {1x1 cell}
    {1x1 cell}

-----------------------------------------------------------------------------

QUESTION 2

I've also got a dataset variable with [x,y,z] data in the following form (also as a cell):

'[4.55566 4.33454 3.45345]' '[3.34534 3.44534 2.23423]'

I'd like to round the values and convert it to something like [4,4,3] [3,3,2]

as a double (maybe?)...

Answer 1

[Peter Perkins]

This

>> nominal(data.sex)
Error using categorical (line 110) Could not compute unique values in A using UNIQUE.
Error in nominal (line 93) b = b@categorical(varargin{:});
Caused by: Error using cell/unique (line 95) Input A must be a cell array of strings.

is probably because you have something other than strings in that particular dataset variable. UNIQUE on a cell array is only defined when all the cells contain strings. For example

>> unique({'m' 'm' 'f' []})
Error using cell/unique (line 86)
Input A must be a cell array of strings.

And this

>> data.subject {1x1 cell} {1x1 cell}

makes me think that you have a cell array each of whose cells contains a 1x1 cell array, which then contains a string. Compare

>> d = dataset({'rachel'; 'tom'},{'s1'; 's2'},{'f'; 'm'},[21; 24],'VarNames',{'subject','subid','sex' 'age'})
d = 
    subject         subid       sex        age
    'rachel'        's1'        'f'        21 
    'tom'           's2'        'm'        24

to

>> d = dataset({{'rachel'}; {'tom'}},{'s1'; 's2'},{'f'; 'm'},[21; 24],'VarNames',{'subject','subid','sex' 'age'})
d = 
    subject           subid       sex        age
    {1x1 cell}        's1'        'f'        21 
    {1x1 cell}        's2'        'm'        24

These

'[4.55566 4.33454 3.45345]' '[3.34534 3.44534 2.23423]'

are strings. I imagine you want numeric values. If all of the observations in this dataset have three values, then you probably want to store the values as a double matrix with three columns. If there can be different lengths of values for different observations, then you want a cell array, each cell of which contains a double vector.

To clean this all up, you probably want to start over, or use some version of cellfun. For example,

data.subject = cellfun(@(c)c{:},data.subject,'UniformOutput',false);
data.x = cell2mat(cellfun(@(s)str2num(s),data.x,'UniformOutput',false))

(where x is the name of the variable you did not provide a name for). Needless to say, this is a bit arcane, because your data are in such a funny form. I don't know how your data got into these funny forms, so I can't really say how to avoid it.

Hope this helps.

Answer 2

[Daniell Algar]

If I understand your data correct, then it's more a question of how you acces the data, than how to convert the data. For instance, take the second question. The following

a= {[4.55566 4.33454 3.45345], [3.34534 3.44534 2.23423]}

would give you the same cell as you have? To perform operands on it, use the cell brackets, as e.g.

a{1}
ans =
    4.5557    4.3345    3.4535

Also, you can access individual elements as usual, ones you've pointed to what part of the cell you are interested in, i.e.

a{1}(1)
ans =
    4.5557

So, to do the rounding you mentioned, just

floor(a{1})
ans =
     4     4     3

The same applies to the data you had in the beginning of your text.

Hope this helps. Good luck