Integrating tolerance level with MATLAB gt(A,B)

Hello!

Is there any good way to use gt(A,B) to determine if A>B with a tolerance level, as in ismembertol(A,B) to determine equality?

Best regards, Nils

Answer 1

[Steven Lord]

A = 1;
B = 1.2;
tol = 0.25;
strictlyGreater = A > B
greaterWithTolerance = A > (B - tol)

[Nils Norlander]

Thank you for your reply!

It doesn't really work properly though.

I have two arrays with 8760 values each, let's say A and B, and a tolerance of 1e-4.

When I try: x = sum( A > (B - tol));

I still get too many values. Any suggestion why?

[Steven Lord]

Impossible to say without seeing a SMALL (not all 8760 values) sample of data that show the problem. Show that data, state what you expect the result to be, and state what you actually see.

[Walter Roberson]

Relative tolerance or absolute?

[Nils Norlander]

So a small sample:

A = [219 219 219 219]; B = [219 219 214.9524 174.8961];

Since the values in B are calculated floating values, the A(1) and A(2) have decimals up to 18 decimals, and strict "<" doesn't do the trick.

Thanks again!

[Steven Lord]

For this example, turn on hex formatting and display A and B. That way we can regenerate the EXACT values of your A and B variables.

format hex
A
B
format % restore default formatting

For those exact A and B vectors, what would you want the output of this to be?

tf = greaterThanWithTolerance(A, B)

If you would expect to call such a function with a user-specified tolerance, what tolerance would you expect to specify to receive your desired output?

tf = greaterThanWithTolerance(A, B, tol)

[Walter Roberson]

What shows up for

fprintf('%.999g\n', B-A)

?

Also please clarify whether you would like the tolerance to be absolute or relative.

Answer 2

[Walter Roberson]

No, there is no good way to use gt() for this purpose: overriding the gt method of double class runs too much risk of breaking things. For example it could break max()