Hi all,
OK, here goes. As someone who was involved with statistics a long time ago, one of the things I would add is that many variables need to be taken into account when analyzing data. For eBird sightings, this means also taking into account the
volume of bird reports.
A report of a bird sighting is a single data entry. It exists (if eBird accepted the record), so you can
suggest a trend, if that bird (data point) is recorded again, by connecting the two entries.
However, when looking at likelihood/probability - and therefore considering trends - the number of potential checklists where it was NOT seen has also to be taken into account. For example: If a person produces one checklist at the same location every day,
the chances of seeing a bird are far higher (seven times) than if only one checklist is produced per week.
In plain English - I would posit that there are more checklists submitted in warmer months than colder ones. More people go birding, and feed data to eBird, when it's nice out, and they believe that they have a greater chance of seeing birds. If that is so,
then data needs to be weighted by a sighting index. Once upon a time I would have known how to do this, but I can't remember how right now. It was a long time ago....
To produce an actual trend, I believe (someone correct me if I am off course here) you have to account for all the reports which do
not show the birds in question, and factor that in to create probability.
Well, you did ask, Gary....
Chris D