Combine vectors of different length into a matrix and perform ANOVA.

Hello,

I have 20 vertical vectors all of different lenghts.

a = 1x876;
b = 1x900;
c = 1x852;
...
y = 1x798;

I'd like to merge them into a single matrix in which each column is one vector (n x 20) and then perform an ANOVA.

Unfortunately a cannot combine vectors of different lenght in one matrix. Is there a way to fill in the empty spaces with a NaN value (probably taking the lenght of the longest one)?

Or maybe there are other methods of which I have no knowledge to perform an ANOVA listing all the variable to be taken into account?

Thank you!

Answer 1

[Adam Danz]

The dimensions you provided indicate that the vectors are row vectors, not column vectors.

If you're performing a 1-ways ANOVA using anova1, you don't have to create a matrix. Use the grouping variable instead. For row vectors,

y = [a,b,c,d];
group = repelem(1:4, 1, [numel(a),numel(b),numel(c),numel(d)])

p = anova1(y,group)

To answer your original question, pad arrays using B = padarray(A,padsize).

[NMarco]

Unfortunately I get this error:

.Error using repelem
In repelem(A, N1, N2, ..., Nn), Nj must be a scalar or a vector of length size(A, j).

I did exactly what you suggested me.

I took all my vectors and stored into y

y = [a,b,c,d];

and then used repelem ;

The first argument is from 1 till the number of vectors (in my case was 1:20), each vector repeated once (1), and then each vector inside numel().

group = repelem(1:4, 1, [numel(a),numel(b),numel(c),numel(d)]);

But I get that error.

[NMarco]

@Adam Danz

update.

I tried to run it using just two of my vectors

>> y = [a, b];
>> group = repelem(1:2, 1, [numel(a), numel(a)]);

and it runs with no errors. So, I'll try to check my syntax, maybe I mispelled something.

CHECKED.

There was a silly syntax mistake. Thank you for your help!

[Adam Danz]

The error is because you forgot the square brackets that group the second input.

repelem(A, N1, N2, ..., Nn)

should probably be

repelem(A, [N1, N2, ..., Nn])

assuming row vector inputs.

[Adam Danz]

Sounds like you figured it out. Glad I could help.