Data de-spiking

On this page, a variant of the median filter can be applied to a data channel of a data table in order to smooth the data. In contrast to median filtering, filtered values are not replaced by a smoothed value but are removed.

The filtering is performed by sliding a window (an interval of a number of consecutive data points) over the data time series. For each time window, the central value of the interval is flagged as a spike if the following condition is met: the ratio of the central value to the mean of three fifths of the sorted (in ascending) values in the interval exceeds the flux ratio specified by the user.

Table and channel selection

Selecting a data table will reload the page with additional selection and input options. Filtering can only be applied to data tables created by the current user.

Next, select a data channel, the interval half width n (each data window will consist of the n values before and after the central data point), and specify a flux ratio value and background percentile. Flux values below the percentile value are considered to be in the background regime and will not be filtered. Set the percentile to zero in order to skip the background check.

Then, press the Run button to start the filtering process. The filtered values of the selected channel will be flagged in the database as NULL values.

Caution: this operation is irreversible!

Last modified on: 2 September 2013.