mhasself / dsmrlib

Down-sampling and multi-resolution support for time-ordered data.

Data Model

Raw data class contains irregularly sampled time-ordered data with time-stamps. Some of the data could be NaN. A measurement is not an instantaneous reading, but rather has some validity over some interval of time -- perhaps half-way to the next-nearest point. This should be described in the raw data container.

First-pass processing of Raw data produces a piece-wise valid interpolator with a particular resolution.

Resampling

The raw data are irregularly sampled, and may have long gaps.

Pass 1: Downsample slightly but produce a regularly sampled vector, to simplify subsequent processing.

Pass 2: For each subsequent down-sampling, process data in contiguous chunks (perhaps with allowances for patching short gaps rather than breaking and restarting).