Collapsing data that are within one integer value of each other

So I've run in to a problem utilizing the zeros generated from using fnzeros. I am able to condense down these zero points, but I need them condensed even more. Basically, there are multiple points in this 1x40 array that are within 1 integer value of each other (19, 20, 21...etc.). They are sometimes groups of 3, but also groups of 2 values which are within 1 integer value of each other. Because of this, if I try to index in steps such as 1:2:length(z) then it will get rid of points that I need. My goal is if there are three values i.e. 19,20,21 then the average is kept (20) and if there are two values then one number is kept i.e. 673,674 will keep 674. I have thought about utilizing logical indexes or possibly sorting the data and then trying to remove the values, but am having trouble figuring out the proper way to do this. I know that since it is a small 1x40 array that I can manually pluck out the data points, but this is very inefficient.

z = 
19  20  21  136  382  631  673  674  760  761  762  877  1124  1372  1414  1415  1501  1502  1503  1618  1865  2113  2155  2156  2242  2243  2244  2360  2606  2854  2896  2897  2983  2984  2985  3100  3347  3595  3637  3638

   inx = 1:length(SA);
  sp = spline(inx,SA);
  z = fnzeros(sp);
  z = round(z(1,:));

Answer 1

[Andrei Bobrov]

n = diff(z(:)) == 1;
lo = [n;0] | [0;n];
nn = [diff([0;n]) == 1;0];
nn(~lo) = 1;
out = ceil(accumarray(cumsum(nn),z(:),[],@mean));

[Michael Jacobson]

Wow, this code works perfectly! I have been reading through the documentation to try and understand it completely. Here's my attempt at explaining it:

-n is the initial logical index that uses diff to track numbers that are within one integer previous of each element

-lo is another logical index that accounts for the fact that the function diff will skip over the first element in the array

-nn = [diff([0;n]) == 1;0]; Not quite sure what exactly is happening with this line. I know a vector is generated from a logical index here.

- This line is setting all nn values in cells which do not equal lo to the value 1

- Reading over the documentation for these functions. accumarray is applying the everage to each subset of elements in z that have identical subscripts in cumsum(nn). cumsum(nn) returns the cumulutive sum of nn. ceil rounds each element to the nearest integer greater than or equal to that element.

Thank you very much Andrei ! If you can help explain your code to me that would be even more helpful. I truly do enjoy learning.