bigger/smaller or equal to for integers not working when numbers are close

I am trying to find integers in a matrix that are within x-percentiles. I am converting the values in my matrix to integers by multiplying and rounding to avoid round errors coming from using floating point numbers. I have used the num2strexact function to confirm that the numbers in my matrix are exact and not approximations. The code is below. In this example, I have set the percentiles to 0 and 100, meaning I should be getting all numbers to meet the conditions I have set. However, when the numbers are close to the percentile values I have set, I am getting some zeroes (as if the conditions were not met). Am I misunderstanding something as how Matlab works (Im new to this). I appreciate any help

PC=round(1000*rand(50,30));

E=round(prctile(PC,0,1));
F=round(prctile(PC,100,1));
ConditionsResults=zeros(50,30,65); 

for n=1:65
    for pc=1:30
        for i=1:50
                    if (PC(i,pc)-E)>=0 & (F-PC(i,pc))>=0
                        ConditionsResults(i,pc,n)=1;
                    else 
                        ConditionsResults(i,pc,n)=0;
                    end;
        end
    end
end

[Walter Roberson]

Possibly to do with the implied all() ? Your test is equivalent to

if all( (PC(i,pc)-E)>=0 & (F-PC(i,p))>=0 )

Is that your intention ?

if all( PC(i,pc) >= E & F >= PC(i,p) )

would be clearer by the way.

[menachom]

No, I want to test all the elements in the matrix individually and then create a matrix that contains '1' if a specific element/value met the condition or '0' if it did not.

For my actual code, I would not have 0 and 100 as the percentile values (eg I could want to set them to 45 and 55 so it finds the values that lie in the middle; I have them at 0 and 100 to test the code since that should give me all 1's).

[Walter Roberson]

Your E and F are both vectors, of length 30. Are you wanting to test PC(i,pc) against the corresponding E entry, E(pc), or are you wanting to test PC(i,pc) against all E entries like you do now?

[menachom]

Gosh, you are right. Thanks a lot. The following fixes the code:

if (PC(i,pc)-E(pc))>=0 & (F(pc)-PC(i,pc))>=0

[Walter Roberson]

That code could be vectorized ;-)

Question: why are you repeating it 65 times when you do not use n as part of the calculations?

[menachom]

Thanks for the input!

The additional loop (65 times) makes sense in the context of the rest of the code (the value of n is used to determine which cutoffs are used)