units/waveform_mean and units/waveform_sd missing sampling rate

[bendichter]

Maybe I am missing something but I do not think it is possible to determine the sampling rate of waveform_mean and waveform_sd in the Units table. To solve this, we could add an optional waveform_sampling_rate attribute to the Units table that is required if these columns are present.

[tjd2002]

I think you’re right. Same goes for the unit field. And do we plan to require that the rate and unit be the same for both waveform_mean and waveform_sd (I think that would be a reasonable requirement). Or, could each of these fields (waveform_mean and waveform_sd) be an electricalSeries object instead of just a bare array?

On Thu, Jun 13, 2019 at 4:01 PM Ben Dichter ***@***.***> wrote: Maybe I am missing something but I do not think it is possible to determine the sampling rate of waveform_mean and waveform_sd in the Units table. To solve this, we could add an optional waveform_sampling_rate attribute to the Units table that is required if these columns are present. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#273?email_source=notifications&email_token=AABLO44OA4ZSYUX3DBPVND3P2LGTRA5CNFSM4HYBBDI2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4GZN23PQ>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AABLO4243LQOGJYHXLYFV5LP2LGTRANCNFSM4HYBBDIQ> .

-- Sent from Gmail Mobile

[rly]

we could add an optional waveform_sampling_rate attribute to the Units table that is required if these columns are present.
Same goes for the unit field.

I agree with adding these fields. The waveform_unit attribute should be fixed to "volts". I also think we should note in the doc of waveform_mean and waveform_sd that those data should be stored in actual converted volts, i.e. not require a conversion factor.

[bendichter]

This has come up twice more in contexts that do not involve the waveforms but where extracellular electrophysiology processing tools expect to have access to the sampling rate of the original data. Because of this, I would prefer we name this field something a little more general, like sampling_rate_of_acquisition. I can't think of a situation where you would need this information and wouldn't also need spikes, so I think it's still fine as a field of the Units table, but might be better off as an optional dataset in /general/extracellular_ephys just in case.

[rly]

I prefer having the sampling rate attached to the group/dataset of the associated data, rather than as a general dataset in general/extracellular_ephys, to reduce ambiguity over what it corresponds to and to prevent data duplication if an ElectricalSeries is provided.

[rly]

After discussion with @oruebel, instead of putting the waveform_sampling_rate and waveform_unit attributes on the Units table, we decided it would be more appropriate to put those attributes on the individual columns that they reference (waveform_mean and waveform_sd). The sampling rate would be duplicated (in almost all cases), but it is more appropriate to have them on the datasets that they refer to in case neither are provided or one is provided without the other.