interp3 returning NaN due to machine precision at bounds

I am having trouble with interp3 returning NaN's at the bounds of the data. Here is a simplified example of my issue...

data = [1 1 1 1 1 1 1 1]
dataRange = 0:0.1:0.7
points = 0:0.1:0.8
interp1(dataRange,data,points(8))

This code returns a NaN, even though points(8)=0.7, which is the upper bound of the data.

points(8)-0.7
ans =
  1.1102e-016

This tells me the points(8) is actually slightly greater than 0.7 by the amount eps/2, making it out of bounds of the data, so interp1 returns NaN. I am using interp3, but interp1 gives the same results. If I instead have points be the same as dataRange, then the problem is gone. But this is a solution I'd like to avoid at this time due to the nature of the code I am writing.

Anyone have any suggestions?

I am running R2010a, Windows 7, 64-bit.

Thanks!

Answer 1

[Teja Muppirala]

You could let it extrapolate outside the region:

interp1(dataRange,data,points(8),[],'extrap')

[Mike]

That works with interp1, but does it work with interp3? I cannot seem to get it to work. I would still like interp3 to return NaN or something else if the point is out of bounds.

I guess my issue has more to do with points(8)=0.70000000000000011102 and not 0.7. Moreover...

points(9) - 0.8

ans =

0

points(8) - 0.7

ans =

1.1102e-016

points(7) - 0.6

ans =

1.1102e-016

points(6) - 0.5

ans =

0

points(5) - 0.4

ans =

0

points(4) - 0.3

ans =

5.5511e-017

points(3) - 0.2

ans =

0

points(2) - 0.1

ans =

0

points(1) - 0.0

ans =

0