How to find inf and sup of a datum not contained in an array ?

Hello all, I have a column vector containing elevations of cross section (x-data are the stations). I have to modify this elevation and I have for example y = 32 mt as output.

In the array I have some elevations that are different from this one and I want to know how to find the index of the first value before and the first value after y = 32 mt (y_new) (with the minimum distance along x from minimum elevation) that are contained in the array and so interpolate the relative new x. 32mt is clearly not contained in the array.

Thanks for reading.

[dpb]

Ya' lost me, sorry...

Give us an example and attach a sample dataset that illustrates what you're after.

Sounds as though interp1 would likely do the trick, but need details to confirm just what it is you're after.

ADDENDUM

interp1 requires the independent variable be unique and not multi-valued which may not cut it if the data are from a FE model coordinates or the like, though.

[Jan]

Some example data would be useful. Is the vector sorted or does it contain unique values? This is more important than the meaning of the values: "elevations of cross section".

Answer 1

[Jan]

x = randi([1, 100], 1, 30);
y = 1 + rand * 99;

[~, index] = min(abs(x - y));
indexBefore = max(1, index - 1);
indexAfter  = indexBefore + 1;

[Virgilio Grigiotti]

Thank you for the answer , I tried in this way , but matlab returns only the nearest index. I can explain better:

for example I have:

y = [40 39.8 39.62 39.32 39.25 39.0 38.8 38.5 37.33 37.6 37.3 32 30 29.5 30 31 32]; % et cetera]'
x = [1 3 5 7 9 12 15 18 20 23 24 30 40 55 72.5 100.12 101.3]; % et cetera]' 
y_new = 37.32;

I want the nearest index in the vector y but having the minimum distance from the lowest y datum.

In your way, it gives me (y(index) = 37.33 ) because is the nearest to y_new and then would give me 38.5.

The result (y(beforeIndex)) must be 37.6 and the second one (y(afterindex)) 37.3 becase this new interpolated evelation is the one with less distance from 29.5.

Thanks.

[dpb]

". I can explain better:"

You're going to have to try again on that front; I can't make heads nor tails of how you come up with the result you do from the above, sorry. :)

Is "beforeIndex" purely positional and dependent upon the unsorted order of y as it is?

Cannot y be sorted and then the x resorted to match up with it?

Just can't follow along and see the purpose here and what is the real algorithm behind the description that defines what the expected answer is so far...

[Virgilio Grigiotti]

Hi, sorry for unclear question, but it's not easy to explain.

Yes, beforeIndex is purely positional and dependent of the unsorted order of y.

Y must not be ordered from smallest to biggest value because the y data must be a hydraulic cross section set with minimum and maximum data in specific positions.

I try to add an image

(x1,y1) and (x2,y2) are the points I'm interested in and I want the position of these ones along the y-data. Then I can interpolate x value of y_new on my own.

In certain way You can sort a part of the datas but you don't know the position in which the y data has linear trend, so y values can be considered unsorted.

[dpb]

This may require iterative solution to test alternatives with the doubly-valued y elevations as you have, depending upon the new y value and just where it is intended for it to go.

It may, in fact, be intractable in general without visual inspection given the shape of the sample curve and the data.

But, for the specific case since it is the last one wanted and the first place that y actually crosses the threshold,

>> ix=find(y<y_new,1);
>> y(ix-1:ix)
ans =
   37.6000   37.3000
>> 

gets the point desired for the particular case. In general is going to be tough methinks...

Here's the given data blown up some to see the detail in the area of interest --

[Virgilio Grigiotti]

it's not valid for all the situations in fact, but thank you for help :)