Bug with mean() on floating point vector elements

Can someone explain this bug?

a = rand(100,1);
mean(a(4), a(6))

Output:

Error using sum
Dimension argument must be a positive integer scalar within indexing range.

Error in mean (line 117)
        y = sum(x, dim, flag)/size(x,dim);

[Stephen23]

"Can someone explain this bug?"

What bug? The fact that you are using mean in a way that is not supported by MATLAB just means that you need to read the documentation.

Answer 1

[Jan]

This is not a bug. Please read the documentation, if you have a problem with a command:

doc mean

The 2nd argument of mean must be the dimension to operate on. Then a floating point number between 0 and 1 will not work. See:

a = magic(3)
mean(a, 1)
mean(a, 2)

You forgot to mention, what you want to calculate. Maybe you mean:

a = rand(100, 1);
mean(a([4,6]))
% or
mean(a(4:6))

[Victor]

OK, then why this code doesn't throw the same error: a = [3,5,2,5,3]'; mean(a(3), a(4))

Understood - it tries to calculate along singleton dimension.

[Steven Lord]

Looking at your a vector, what is a(4)? It's 5. Can you take the mean of a two-dimensional array in the fifth dimension? Sure. While a(3) doesn't look like it has a fifth dimension, if you ask for the size of a(3) in the fifth dimension what does MATLAB return?

size(a(3), 5) % returns 1

See this Answer for some discussion of trailing singleton dimensions.

The reason your original attempt didn't work is that arrays in MATLAB don't have (for example) a 0.25th dimension.

[Stephen23]

@Viktor: it does not try to calculate along the singleton dimension, it actually does calculate along the singleton dimension.