How important is it to provide historic or time-series data?


#1

Data publishers often only publish a snapshot of the latest data. Sometimes it’s useful to keep a history of these releases so data can be compared over-time.

How important is it to you that historic or time-series open data is available?


#2

I’ve found time series data extremely valuable for trend analysis - provided the data is captured at regular intervals, using the same methodology if possible, and where any differences in sampling or methodology are clearly communicated as part of the context of the data, so it can be normalised where required.

Going backwards in time, government should publish as many existent datasets in a time series as possible - potentially requesting assistance from the community where earlier data exists in non-machine-readable formats.

Going forward, ultimately this is impacted by the collection capability of agencies, which is impacted by their internal mandate and need for the data and commensurate funding.

I don’t believe agencies should feel constrained to keep capturing data where they don’t have capability - though communities may have strong views on this. However wherever possible retained data should be published to provide a timeline (with collection context) and data series should continue to be captured and published where value can be demonstrated.

More of a concern for me is how we compare datasets collected by different states and territories to provide a national view of topics. There’s currently far too many cases of different methodologies, time periods and data quality discrepencies which make it very hard to build a national picture, or compare specific states/territories in order to derive deep insights or provide an attractive (national) consumable service.


#3

Agreed with Craig. Time series data is extremely valuable and allows some very interesting data analytics. But there does need to be standardisation - at least between different time periods of the same dataset, and also between datasets/states/territories too if possible.


#4