On 18 Apr 2011, at 16:46, Toby Murray wrote: > On Mon, Apr 18, 2011 at 10:25 AM, M∡rtin Koppenhoefer > <dieterdre...@gmail.com> wrote: >> 2011/4/18 David Murn <da...@incanberra.com.au>: >>> If you want to represent these important figures in statistics, can you >>> at least use a common scale to avoid distorting peoples views of the >>> figures? Using deceptive graphing methods was a trick we were taught >>> back in school as a child. It doesnt make your figures look any better, >> >> >> It makes them readable. If you used the same scale you won't see the >> handful of no-votes against the 10000 yes-votes. > > Yes, this is why I used a different axis for both values. Otherwise > the "accept" would be a straight line across the top and the "decline" > would be a straight line across the bottom of the graph. Not very > useful. > > I am using zabbix to make the graphs. Like I said, it is targeted at > system monitoring, not statistical analysis. Hence, the scales change > based on the available data to maximize the viewability of the data. > > If someone wants, I might be able to produce a data dump so you can > make your own graphs. Zabbix stores it as a timestamp and a value in a > mysql database.
While I agree that there is a problem with the no votes disapearing if you show the whole graph, it would be useful to show the same *range* on each scale. I.e., as we are currently showing 10300 - 10900 on the yes scale, show 0 to 600 on the no scale. This will give a much clearer indication of the trend. Bob _______________________________________________ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk