In moving transform with overlapping windows, it should be possible (as an option) to exclude the first n rows from the output, with n the size of the window - after their values have been used to calculate the outputs for rows [n+1 .. 2n]. The first n rows have less than n predecessors, therefore they are calculated differently from rows n+1 onwards. It is currently even not specified how their output values are determined. This should also be made explicit, in case the suggested option is not selected.
Actual behavior
I assume the of the first n rows is based on applying the aggregation function to each row itself and its (less than n) predecessors. It might even be that leading zeroes are used - I haven't checked
Steps to reproduce the behavior
Apply Moving Transform (overlapping windows) to a time-series dataset and check the output.
Additional info (worksheets, data, screenshots, ...)
Timeseries version
0.3.10
Orange version
3.28
Expected behavior
In moving transform with overlapping windows, it should be possible (as an option) to exclude the first n rows from the output, with n the size of the window - after their values have been used to calculate the outputs for rows [n+1 .. 2n]. The first n rows have less than n predecessors, therefore they are calculated differently from rows n+1 onwards. It is currently even not specified how their output values are determined. This should also be made explicit, in case the suggested option is not selected.
Actual behavior
I assume the of the first n rows is based on applying the aggregation function to each row itself and its (less than n) predecessors. It might even be that leading zeroes are used - I haven't checked
Steps to reproduce the behavior
Apply Moving Transform (overlapping windows) to a time-series dataset and check the output.
Additional info (worksheets, data, screenshots, ...)