flag_outliers()

std::vector< size_t > MayaFlux::Kinesis::Stochastic::Estimate::flag_outliers ( std::span< const double >  samples, double  threshold_mad = 3.0  )

staticnoexcept

Flags samples whose deviation from the median exceeds a threshold.

Parameters

samples	Values to analyze
threshold_mad	Number of scaled-MAD units beyond which a sample is flagged, typically 2.5 to 3.5 for physical sensor data

Returns

Indices into samples considered outliers

Shape-aware in the sense that a single wild sample and a run of several consistent high samples are distinguished by the caller inspecting which indices come back, not by this function alone: an isolated flagged index surrounded by unflagged ones reads differently from several consecutive flagged indices, and that read is left to the caller since the correct interpretation depends on domain (a genuine fast transition vs. a dropout burst).

Definition at line 247 of file Estimate.cpp.

{
    std::vector<size_t> result;
    if (samples.size() < 2)
        return result;
 
    std::vector<double> sorted(samples.begin(), samples.end());
    std::ranges::sort(sorted);
    const double median = sorted[sorted.size() / 2];
    const double scaled_mad = median_absolute_deviation(samples);
 
    if (scaled_mad < 1e-12)
        return result;
 
    for (size_t i = 0; i < samples.size(); ++i) {
        const double dev = std::abs(samples[i] - median) / scaled_mad;
        if (dev > threshold_mad)
            result.push_back(i);
    }
 
    return result;
}