How to find these rows in a dataset directly before a sequence of NaNs?

Hey people. I have a huge dataset which looks as follows

   x y z p
   2 4 1 20
   2 4 2 1
   2 4 3 NaN
   2 4 4 NaN
   2 4 5 NaN 
   3 5 1 5
   3 5 2 1
   3 5 3 52
   3 5 4 22
   3 5 5 NaN
   6 3 1 10
   6 3 2 4
   6 3 3 1
   6 3 4 NaN
   6 3 5 NaN
  (...)

As u see the dataset is sorted according to the first three columns. I would like to extract only the lines/rows which are directly above the NaN. I find a way to remove all the lines containing a NaN by using an index.

p(isnan(p))= -999;
index = 1:size(p,1);
index = index(p > -999);
x = x(index);
y = y(index);
z = z(index);
p = p(index);

But I have not really a clue how to remove the remaining lines. Does someone have an idea/hint?

so the solution for the example would be:

     2 4 2 1
     3 5 4 22
     6 3 3 1

Thanks in advance! P

Answer 1

[the cyclist]

% Find rows that are not NaN themselves, but are followed by a NaN line. (Not sure how you wanted to handle the final row. This code keeps it, if it is not NaN.)
indexToKeep = [not(isnan(p)) & isnan([p(2:end); NaN])];

x = x(indexToKeep)
y = y(indexToKeep)
z = z(indexToKeep)
p = p(indexToKeep)

Answer 2

[Joseph Cheng]

so you can create something like this:

sampdata = [2 4 1 20
   2 4 2 1
   2 4 3 NaN
   2 4 4 NaN
   2 4 5 NaN 
   3 5 1 5
   3 5 2 1
   3 5 3 52
   3 5 4 22
   3 5 5 NaN
   6 3 1 10
   6 3 2 4
   6 3 3 1
   6 3 4 NaN
   6 3 5 NaN];

pmask = isnan(sampdata(:,4));
dpmask = diff(pmask);
 % index location of a number to Nan transition ==1 
 % index location of a Nan to a number transition ==-1 
 % 0's are continuous types
outputdata = sampdata(dpmask==1,:)