Filtering timeseries with NaN

I am filtering out data due to the degree of Standard deviation. Once I determine which rows of data are good, I good back to the orginal data set and try to NaN the rows which are 'bad'.

threshold = 0.0013;
filtRows = find(relSD1 < threshold);

% plot(time, relSD1, '.-');
figure(6)
%plot figure - note that you need to rename my 'time' for your own time vector name
plot(time, f_fluxxcorredit,'.-k');
hold on;
plot(time(filtRows), f_fluxxcorredit(filtRows),'.r');

legend('all data', 'filtered data'); 
xlabel('Time, hrs');
ylabel('Flux mmol m^{-2} day^{-1}');
shg

How would I get a vector like f_fluxxcorredit (201 points) but with the bad data NaN? Filtrow is showing the good points 112 of them and the others should be NaN of f_fluxxcorredit.

Answer 1

[jonas]

It's easier to use logical indexing. If good data lies below a certain threshold, then bad data lies above that threshold. Switch the sign and use the condition for logical indexing.

This should work:

threshold = 0.0013;

f_fluxxcorredit(relSD1 > threshold)=NaN;

[Rebecca Ellis]

Great this works, thank you again.

Answer 2

[Peter Perkins]

You might take a look at isoutlier and filloutliers.