Skipping time windows in outlier test

I have a large dataset of 1second time-series data in a timetable. In the dataset I also have a binary flag for when there is a filter in place for backgrounding the measurement. I would like to detect outliers for the signal measurement to remove erroneous points before doing any calculations on the data. Is there a simple way to incorporate the filter flag using isoutlier() so that I can ignore those points without the moving median incorporating the last sample time?

The data set has ~10 million points which makes me want to avoid looping through.

[Image Analyst]

Perhaps rmoutliers or isoutlier?

Exactly which times should be removed?

s = load('question_TT.mat')

s = struct with fields:
    question_TT: [3000×11 timetable]

tt = s.question_TT;
plot(tt.Time, 'b.-')
grid on;
xlabel('index');
ylabel('time')

[Poison Idea fan]

I've used isoutlier on variables when "filter_state" == 0 but the timing is discreet. So when I use isoutlier, it's using the MAD from the previous chunk of time when filter_state==0.

[Image Analyst]

You will need to identify the discontinuities and just set the "bad" MAD value at the discontinuities to the next valid "good" value.

[Poison Idea fan]

I see. Thanks.

[Poison Idea fan]

So, what I found was that I was not re-synchronizing the outlier flag output from the isoutlier function with the initial flag I used for sample points rather than background points. Mistake by me.

Thanks for your response.