@lgolowich has been adding some nice time series support to Data Explorer.
The current approach to configuration, however, makes this hard to adopt for AMP PD.
The current configuration option is to specify a single time_series_column across the entire dataset. Given that AMP PD has a visit_month in every table, this would instantly turn every table/facet into time series data if we set the time_series_column.
This doesn't always make sense. In many cases, one really wants to see the data in aggregate and in general there is just a single record (at most) per individual.
Some examples:
Demographics
Caffeine_history
Family_History_PD
In these cases, there is generally at most one record per individual, the visit month is a useful detail to have in the data, but when viewed through Data Explorer, it is uninteresting and unhelpful to have the data separated by visit month. That one contributing study collected data at a screening visit (-1), another study collected data at baseline (0), and another collected data at a follow-up visit (0.5) may be interesting, but probably not the way that I want to view the data.
I really think we will need the ability to either:
Turn on time-series on a per-table basis, OR
Multiple views of the data that a user can toggle (aggregate vs. time-series)
I think that the latter would be really cool, frankly - it would give users the most control over the way that they view the data. It seemed like the former would be easier to implement and would give data publishers more control over the "out of the box" experience for their users.
@lgolowich has been adding some nice time series support to Data Explorer.
The current approach to configuration, however, makes this hard to adopt for AMP PD. The current configuration option is to specify a single
time_series_columnacross the entire dataset. Given that AMP PD has avisit_monthin every table, this would instantly turn every table/facet into time series data if we set thetime_series_column.This doesn't always make sense. In many cases, one really wants to see the data in aggregate and in general there is just a single record (at most) per individual.
Some examples:
In these cases, there is generally at most one record per individual, the visit month is a useful detail to have in the data, but when viewed through Data Explorer, it is uninteresting and unhelpful to have the data separated by visit month. That one contributing study collected data at a screening visit (-1), another study collected data at baseline (0), and another collected data at a follow-up visit (0.5) may be interesting, but probably not the way that I want to view the data.
I really think we will need the ability to either:
I think that the latter would be really cool, frankly - it would give users the most control over the way that they view the data. It seemed like the former would be easier to implement and would give data publishers more control over the "out of the box" experience for their users.