OK, I am not sure I understand the issue correctly, so I'll just throw out some notes:
* I do not count the pageviews. They are counted by the Wikimedia Foundation, I just use them as-is. * The official definition of page views seems to be at https://wikitech.wikimedia.org/wiki/Analytics/Pageviews * AFAIK, the data before 2015-12 used only desktop views; the new one uses also mobile, but removes "views" by bots and crawlers. * Not sure what the "bounce" feature is; the preview thing in the mobile app? Or the MediaViewer? (I believe page views count neither of those) Finally, a "view" in the baglama2 tool means an image was included on a page that a human loaded in his/her browser. I have no data if that image was actually on the screen ("below the fold", no scrolling), but since "important" images in an article tend to be near the top, I just count the page view. Hope that helps, Magnus On Mon, Jan 18, 2016 at 2:19 PM Federico Leva (Nemo) <nemow...@gmail.com> wrote: > Hans Muller, 18/01/2016 14:52: > > As far as i can see, the API does not answer the bounce question. > > The API talks of pageviews > https://wikitech.wikimedia.org/wiki/pageviews_API ; your question is > IMHO the wrong one, i.e. "how can we estimate how many pageviews are > real?". The real question is "how many times are the files actually seen?". > > To answer the real question, you have to use mediacounts instead: > https://wikitech.wikimedia.org/wiki/Analytics/Data/Mediacounts > You can see an example at > http://wiki.wikimedia.it/wiki/File:2015-10-beic-counts.ods > The results are very similar to what baglama yields, for this specific > case (BEIC media). > > Nemo > > _______________________________________________ > Glamtools mailing list > Glamtools@lists.wikimedia.org > https://lists.wikimedia.org/mailman/listinfo/glamtools >
_______________________________________________ Glamtools mailing list Glamtools@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/glamtools