How to replace array elements in specified indexes with another array?

Good morning/afternoon/evening, ladies and gentlemen.

A bit of introduction to the problem for your convenience and ease of understanding:

• Imagine I have an array "A", which is 10x1 double. The values are [1; 2; 3; 4; ... 10] with indexes 1, 2, 3 ... 10, respectively.
• To find outliers and I made following array: TF = isoutlier(A). Assume the TF is 5x1 double and it has detected outliers in array "A" with indexes/positions of outliers of [3, 4, 6, 8, 9].

Question:

1. How do I delete the array "A" values in indexes mentioned by array "TF" without completely removing indexes - for example A = [1, 2, NaN, NaN, 5, NaN, 7, NaN, NaN, 10]?
2. And how do I replace the aforementioned "NaN" values in array "A" with mean values of first available neigbouring values? The example output I'm looking for is following: A = [1, 2, ((2+5)/2), ((2+((2+5)/2))/2), 5, ((5+7)/2), 7, ((7+10)/2), ((7+((7+10)/2))/2), 10]?
3. Since I foreshadow that the outlier can be at the end of array, how do I write this into rule explained in question #2? Example A = [1, 2, ... ((7+???)/2)]

I apologise beforehand if the questions are dumb, I am new to this, hence I'm dumb myself. Thank you, kindly, for your assistance and advises.

Best regards,

Mekan

Answer 1

[DGM]

It can be done, something like this:

https://www.mathworks.com/matlabcentral/answers/829183-how-to-remove-outlier-of-time-series-signal-data-by-using-cubic-spline-interpolation

... but you could also use filloutliers(). I'm assuming you mean to interpolate linearly between the endpoints of the gaps instead of just using a constant value in between.

t = 1:10;
x = 1:10;
oloc = [3 4 6 8 9];
x(oloc) = NaN; % just for example

mask = false(size(x)); % convert oloc into a logical mask
mask(oloc) = true;
xrfo = filloutliers(x,'linear','outlierlocations',mask);

plot(t,xrfo,'g*-'); hold on; grid on
plot(t,x,'ko')

In practice, it wouldn't be necessary to pre-fill the outlier positions with NaN. Removeoutliers() will ignore those values anyway.

Or alternatively, you can use interp1() for a case as simple as this.

clf
t = 1:10;
x = 1:10;
oloc = [3 4 6 8 9];

mask = false(size(x));
mask(oloc) = true;
ts = t(~mask);
xs = x(~mask);

xrfo = interp1(ts,xs,t,'linear','extrap');

plot(t,xrfo,'g*-'); hold on; grid on
plot(ts,xs,'ko')

[Mekan Nuvryyev]

Thank you deeply, DGM!

I have used your answer to do deeper research on Matlab functions and I have found solution! It was way simpler than I have anticipated. As a matter of fact it was so simple, I now am embarassed of my questions. I will show solution below in hopes that it will be useful for someone else:

[TF,L,U,C] = isoutlier(A,'quartiles');                           % A - array needed where outliers need to be filtered and filled
                                                                 % TF - logical array of indexes of detected outliers
                                                                 % L/U/C - Lower/Upper/Centre boundaries/values of detection method                          
[A_filtered,TF,L1,U1,C1]=filloutliers(A,'linear','quartiles');   % L1/U1/C1 are created separately from L/U/C values to double confirm that the L1/U1/C1 values are same as L/U/C

Thank you, again, DGM, for your kind assistance!