Interp1 Function gives back NaN in second to last datapoint

Hi everyone! Im trying to solve an issue for my graduation where i need to use interpolation. I have a dataset that i can use to simulate a full movement with help of interpolation. Somehow when i use the Interp1 function the second to last datapoint is NaN. For plotting my data this isn't a issue, because i dont use this point within my plots, but to determinate the Root Mean Squared Error i need this number to compare it with the full movement. The code i use for creating the interpolated data is as followed:

FE = a(571:603);

x = x(169:201);

Frames = [FE(1) FE(17) FE(33)];

3Int = [x(1) x(17) x(33)];

inter = interp1(Frames, 3Int, FE,'linear')

Be aware that i know the points of the complete movement. I just want to know if using an interpolation method gives back a good representation of the original dataset. I am aware that of a 33 point data set a 3 point method of interpolation isnt ideal, but it also occurs when i use more point to interpolate.

The data of FE and x are as followed:

FE:

37.8570

36.1267

34.1324

32.0503

30.2909

29.0938

26.7495

25.3398

23.2135

21.6724

19.8791

17.1966

15.7437

13.4649

11.8510

8.6762

5.8552

4.3892

2.7858

0.9258

-2.8756

-6.6230

-9.2144

-12.0599

-14.3876

-16.7708

-19.1892

-21.1271

-23.5877

-25.9523

-27.8837

-28.8928

-28.8405

x:

-0.2599

-0.2422

-0.3031

-0.3396

-0.4032

-0.4599

-0.5259

-0.6096

-0.6955

-0.7171

-0.7418

-0.7406

-0.6971

-0.6624

-0.5449

-0.4357

-0.2800

-0.1954

-0.1601

-0.0478

0.0479

0.1878

0.3040

0.4008

0.5121

0.6089

0.7325

0.8684

0.9627

1.0562

1.0962

1.0923

1.1059

Thanks in advance!

Answer 1

[Bruno Luong]

MATLAB interp1 by default gives NaN for query points falling outside your source points (first and last).

Use 'extrap' option (not recommended unless you know what you are doing)

[Sven Dobbe]

Thanks! very clear

Answer 2

[Torsten]

Frames = [FE(33) FE(17) FE(1)];
3Int = [x(33) x(17) x(1)];

instead of

Frames = [FE(1) FE(17) FE(33)];
3Int = [x(1) x(17) x(33)];

The data points must be distinct and in ascending order.

You will also use a "sort" on your full vector FE if you want to use it for interpolation.

[Bruno Luong]

"The data points must be distinct and in ascending order"

descending order works fine also.

>> Frames = [FE(1) FE(17) FE(33)]

Frames =

37.8570    5.8552  -28.8405

>> issorted(Frames,'monotonic')

ans =

logical

1

[Torsten]

@Sven Dobbe

Tast two data points in the long FE vector are not in order.

This gives the NaN values because of the extrapolation problem.

[Sven Dobbe]

Yes i indeed can see that this causes an issue. The problem is that i cant rearange this order to sort the data. This gives an unrealistic movement that doesnt exist. This is probably a one in a 100 cases where this happens. This obviously needs to be adressed later on.