histc problem in counting in particular bin

Hello

For data = 0.1:0.02:0.3 and bins = 0.1:0.04:0.3. When I use histc(data, bins), I get histogram frequency = 2 2 2 2 2 1. which is correct.

But when I change bins upto 0.4, I mean bins = 0.1:0.04:0.4, then the histogram frequency = 2 2 3 1 2 1 0 0, which is not correct. It incorrectly counting upper bound in 3rd bin and missing lower bound in 4th bin.

I appreciate any help.

Answer 1

[Matt J]

You are assuming that the bin boundaries are calculated with infinite precision, which is impossible. Likewise your data is also not computed in perfect increments of 0.04.

[Matt J]

One remedy would be to scale everything up to integers,

data = 0.1:0.02:0.3;
bins = 0.1:0.04:0.3;

histc(round(data*100), round(bins*100))

ans = 1×6

2 2 2 2 2 1

[Dammar Badu]

Thank you Matt.

Scaling to integers really works. It is very helpful.

But I still do not understand why bin boundaries are working perfectly in some bins and not in other.

Any way problem is solved with your solution. Thanks again.

Dammar

[Image Analyst]

Then read the FAQ entry:

https://matlab.fandom.com/wiki/FAQ#Why_is_0.3_-_0.2_-_0.1_.28or_similar.29_not_equal_to_zero.3F

[Matt J]

But I still do not understand why bin boundaries are working perfectly in some bins and not in other.

The floating point calculation errors are essentially random. You can think of it as though small random noise was added to both bins and data. Therefore if a data point lies very close to a bin boundary, there is no way to predict in which bin it will be counted.

[Walter Roberson]

The process is not actually random, but it is complicated . If you have two different ways of calculating the same mathematical quantity, then if there are any differences, you might not get the same result for the two ways.

[Dammar Badu]

Thank you Matt, Walter and Image Analyst for clarification. It is very helpful.