In another post where I asked about discrepancies between exported data for 15 minutes versus hourly time bins, I learned that there is an implicit adjustment for uptime that is applied to the hourly counts displayed on the public web page (but not the 15 minute counts in exported tables).
So my first question: is an implicit uptime correction also applied for hourly ‘BicycleTotal_10Modes_’ counts in the exported spreadsheet files? or do the tabulated values represent uncorrected raw counts?
Also, it was mentioned that uptime fractions less than one reflect 15-minute bins that have failed to transfer to the server. Would it thus be correct to assume that the meaningful values for hourly uptime are effectively constrained to zero, one quarter, a half, and one? The many decimal places of precision could imply otherwise but it seems like this might just be an artifact of the export. If the fraction just indicates lost uploads, uptime has no relevance in the context of a 15-minute time bin (since it either uploads or it doesn’t).
I am doing offline analyses of exported hourly count data and want to make sure that I do not apply a redundant adjustment for uptime.
I realise this was posted over the holidays so it was some time ago, but I wonder whether you got a reply to this directly from our support, or if this is still something you are looking into and need support for?
yes, a correction is applied to the reported data for ANY period in our data, both in the reports and data from the API and that is why we also publish the uptime data used for the correction
that is basically true
This is not correct. While it may be the case that an upload failure means we have lost the full 15 minutes of data for some reason (e.g. a temporary power loss at the time of the upload), it is also possible that the device reboots during that 15 minute session, and therefore loses a fraction of that time. This will only usually be a few seconds, but we apply a correction in any case, for consistency. You would then have data for [15 + 15 + 15 +14] minutes for a particular hour, and therefore an uptime that is less than 1.
Thank you - this is exactly the information I needed.
So it seems that there is never any need to apply uptime correction to the count values in exported spreadsheets. If I understand correctly now the uptime value is simply provided on an informational basis.
Just to close the loop fully, the initial question arose because I noted a discrepancy between count totals based on the 15 minute export data and the web data, which is apparently based on 60 minute bins. This was attributed to the uptime correction being applied for the web data (and apparently for the 60 minute export data) but not for the 15 minute data. Based on this, it would seem that the 15 minute data might be an exception to the statement that correction is applied to any period. Sorry to belabour the point - just want to make sure I understand. Thx