0

Perhaps this is a bit redundant but still, it bothers me a bit. Why do I need to turn a vector into an integer class to get the element-wise comparison correct?

# Finding a linear combination of v, in terms of u1,u2,u3
v=c(3,7,-4)

u1=c(1,2,3)
u2=c(2,3,7)
u3=c(3,5,6)

# L. System AX=B
A <- cbind(u1,u2,u3)
B <- v
k <- solve(A,B)
k
## Therefore
u <- k[1]*u1+k[2]*u2+k[3]*u3

# First elemement FALSE?
v==u
identical(v,u)
all(v==u)
v %in% u

#Both numeric
is.numeric(v)==is.numeric(u)
u <- as.integer(u)

# The elemnt-wise comparison is now correct, identical FALSE, because they are a different class.
v==u
identical(v,u)
all(v==u)
v %in% u
Mario GS
  • 859
  • 8
  • 22
  • 1
    This is a numerical precision issue. Try `print(u, digits=20)` after it is created in `u <- k[1]*u1+k[2]*u2+k[3]*u3`. – lmo Apr 01 '17 at 13:44
  • 1
    `u-v = [1] 1.776357e-15 0.000000e+00 0.000000e+00` – Ben Bolker Apr 01 '17 at 13:45
  • 1
    The general recommendation when working in these cases is to use `all.equal`, which has a tolerance argument. Try `all.equal(u,v)`. – lmo Apr 01 '17 at 13:50
  • 1
    Ahh, I suspected that... Thanks for pointing that out, I didn't know about `all.equal`. – Mario GS Apr 01 '17 at 15:09

0 Answers0