Mean based on positive and negative values of a field

I have a structure S (10000 X 1) with 10 fields. I want to calculate (two) means and standard deviations of one particluar field f1. I have condition on field f2.

mean_1 and SD_1: consider the values of f1 (in calculation of mean and SD) if, f2 is positive.

mean_2 and SD_2: consider the values of f1 (in calculation of mean and SD) if, f2 is negative.

Any smart way to do this?

Answer 1

[Rik]

Logical indexing:

L=[S.f2]>0;

pos_mean=mean([S( L).f1]);
neg_mean=mean([S(~L).f1]);

[SS]

Thanks Rik. It works well. Just for the information, what should be the syntax if it is a structure array?

[Rik]

I don't understand your question. This code already is for a struct array. Or do you mean a struct array of size n-by-m with both n and m greater than 1?

[SS]

If S is 1 X 20000. Yes, that is what I meant.

[Rik]

This should also work for that shape. There might be some unexpected results if the contents of your fields are non-scalar.

What I meant is e.g. a struct array of size 20 x 30.

[SS]

S=struct;

S(1).f1=[2,3,4,5,6,7,8,10,14,16,18,20];

S(1).f2=[-3,6,9,-12,15,18,-21,24,27,-30,33,-36];

S(2).f1=[4,5,6,7,8];

S(2).f2=[-0.1, 4,0.8,-3,4];

Something like this. When, I tried to adapt the above syntax to this structure array it gives the follwing error:

"The logical indices contain a true value outside of the array bounds."

[Rik]

As I mentioned, non-scalar structs complicate it a bit.

all_f1=[S.f1];
all_f2=[S.f2];

L=all_f2>0;
pos_mean=mean(all_f1( L));
neg_mean=mean(all_f1(~L));

Note that the concatenation will probably not work if your fields have a different direction than your struct array. So this should work, but a 100X1 struct array with 1Xn fields will not.

[SS]

Great, thanks.