Regarding removal of low values within time series

I have a time series data (see attached data_qn.xls) that consists of values that are clearly lower than the rest of the observations. How could I remove these values in MATLAB?

I considered running a sliding window function that returns the lowest values, but it seem impractical because the window size may be different. I also tried something like the code below, but the indices are relative to its own window array and that complicates things.

for i = 1:length(data)-window_sz % window_sz = 20 arbitrarily
  window_range = data(i:i+window_sz-1,:);
  [M I] = min(window_range);
  del_rows(i,1)=I;
end

Any idea?

Thank you Ben

Answer 1

[KSSV]

data = xlsread('data_qn.xls') ;
data(data<4) = [] ;    % remove values less then 4

[BenL]

I pondered over thresholding but I realised that not all of my datasets can perform thresholding at a consistent value, even though the 'line' of lowest values exist.