k=-19.252747
slope=-19.252747
-->-19.252747--19.252747
ans =
0.
-->k-slope
ans =
2.132D-14
So, what is the difference?
Note:
k and slope calculated by;
bi=[0.01,0.02,0.03,0.04,0.05,0.06;
-0.19,-0.39,-0.57,-0.77,-0.96,-1.16]
k=bi(2,:)/bi(1,:)
[m,n]=reglin(bi(1,:),bi(2,:));
fity=m*bi(1,:)+n;
slope=tan(atan(fity/bi(1,:)))
It is nothing specific here as results of k and slope shown here are 8 place accurate while their difference is shown in double precision to 15-16 place accuracy and is nearly 0 or may say they differ in values at 11th place onwards.
Values in double precision have only roughly 16 decimal digits of precision. For your example there are already two digits in front of the decimal point which leaves you with roughly 14 digits after the decimal point (actually, it is more like 13.955 digits). After so many computations for 'slope', especially including tan and atan, there will definitely be a loss in precision. Since the computer can never have an exact representation of the numbers every computation will induce a tiny error. With more chained computations the error will propagate.
This means that with the result 2.132E-14 the difference in the two values 'k' and 'slope' is only in the last one or two bits. You cannot expect more from a computer.
- Badges
- Science topic