Resolution: Won't Fix
Affects Version/s: 0.9.4
Fix Version/s: 0.9.9
NOTE: This will require timeseries datatypes to know their sample-rate, etc.
I'd suggest changing the Timeseries visualiser to allow users to select the
length of the display window in time units (ms, s, min, hours, etc.)
The display window should then have a fixed resolution – for my setup a
sensible value would be around 1000, it possibly make sense to have this
user setable, maybe ~(250, 500, 1000, 2000, 4000).
The set of selectable lengths and resolutions should be constrained by the
selected timeseries. To simplify implementation the window length/resolution
combinations should be restricted so as to produce integer multiples of the
data's period. This would mean we only need to subsample the original
data, avoiding the need to deal with interpolation.
As the set of "sensible" window lengths can be quite large, perhaps a slider
type setter would be appropriate. For example, for M/EEG I can imagine the
following values being a "sensible" set that would cover most use-cases:
(100ms, 500ms, 1s, 2s, 5s, 10s, 20s, 30s, 1min, 2min, 5min, 10min, 20min, 30min,
1hour, 2hours, 5hours, 10hours, 24hours)
The shortest values covering the highest frequencies common to M/EEG data and
the longest values for clinical sleep and epilepsy studies where patients can
wear recording equipment for days...
For TVB timeseries datatypes we can specify sensible time lengths, for example,
for BOLD anything less than 1 minute doesn't make much sense – so you'd exclude
the first 8, or maybe 10, values in the list above – and people basically never
lie in a MRI scanner for more than a few hours – so, the last few could be
dropped as well...
There should also be logic that says that for the specified timeseries length
in time units if the data contains less data-points than the specified
resolution then the length and resolution should be automagically reduced to
the actual data length and number of data-points, so the tiemseries fills the