Prev: Proposed Motion of No Confidence in URCM Moderation
Next: Ok cyclists - is this reasonable behaviour?
From: Albert T Cone on 31 Mar 2010 10:37 boltar2003(a)boltar.world wrote: > On Wed, 31 Mar 2010 10:18:03 +0100 > Albert T Cone <a.k.kirby(a)durham.ac.uk> wrote: >>> Have you read about the "biased" presentation, including the fact >>> that Dr Walker chose to present his results graphically with a >>> non-zero origin for his axis showing Main Overtaking Proximity which >>> gives a false representation of the results. >> That is a perfectly proper way to resent data with a bias - it maximises >> the resolution with which the data can be read. It only gives a false >> impression of the data if the reader doesn't pay attention to the axes; >> that is not the fault of the author. > > On the contrary , a lot of authors do it deliberately in the hope that > many people WONT look at the axes and even if they do the psychological > effect of the graph can still bias them towards what the author wants > them to belief. You can see these dodgy graphs everywhere from sales charts > to government statistics to arguments for and against climate change. If we are talking about a scientific publication, as I understood us to be, then the proper way to present the data is that which lets it be read most accurately; in this case with axes offset. Anyone of a scientific grounding *should* read the axes carefully. I would agree that in a presentation for the benefit of the general public, who might be used to axes crossing at (0,0), then it could be a source of confusion, and can readily believe that is used deliberately in marketing and similar chicanery.
From: Nick Finnigan on 31 Mar 2010 12:33 Albert T Cone wrote: > > If we are talking about a scientific publication, as I understood us to > be, then the proper way to present the data is that which lets it be > read most accurately; in this case with axes offset. Anyone of a > scientific grounding *should* read the axes carefully. It is read most accurately with the Origin at zero. If the variation can not be seen with that presentation, it is not important.
From: Patrick Gosling on 31 Mar 2010 13:41 In article <hovtgk$5js$1(a)news.eternal-september.org>, Nick Finnigan <nix(a)genie.co.uk> wrote: > It is read most accurately with the Origin at zero. If the variation can >not be seen with that presentation, it is not important. Oh the joy of cargo-culted rule-making. There are plenty of situations where "zero" is an entirely inappropriate place for an origin. Do you suppose that an investigation into the dependency of a mechanical system on the current barometric pressure will be improved by an insistence on not having an origin somewhere near 1000 mbar? Do you perhaps think that a plot of some property against temperature should definitely have its origin at 0'C ? If so, why? If not, why not? If I find that my graph lacks clarity when the origin for parameter Y is at zero, should I perhaps instead a plot and label it as a graph of parameter Z (which happens to be Y + a constant)? -patrick.
From: Nick Finnigan on 31 Mar 2010 13:54 Patrick Gosling wrote: > In article <hovtgk$5js$1(a)news.eternal-september.org>, > Nick Finnigan <nix(a)genie.co.uk> wrote: >> It is read most accurately with the Origin at zero. If the variation can >> not be seen with that presentation, it is not important. > > Oh the joy of cargo-culted rule-making. > > There are plenty of situations where "zero" is an entirely inappropriate > place for an origin. > > Do you suppose that an investigation into the dependency of a mechanical > system on the current barometric pressure will be improved by an > insistence on not having an origin somewhere near 1000 mbar? Yes. > Do you perhaps think that a plot of some property against temperature > should definitely have its origin at 0'C ? If so, why? If not, why not? 0C is a false origin, not 0K. > If I find that my graph lacks clarity when the origin for parameter Y > is at zero, should I perhaps instead a plot and label it as a graph of > parameter Z (which happens to be Y + a constant)? Where there is a clear zero (e.g. risk or distance) it should be used.
From: Robin Stevens on 31 Mar 2010 14:28
In cam.transport Nick Finnigan <nix(a)genie.co.uk> wrote: > Patrick Gosling wrote: > > Nick Finnigan <nix(a)genie.co.uk> wrote: > >> It is read most accurately with the Origin at zero. If the variation can > >> not be seen with that presentation, it is not important. > > Do you suppose that an investigation into the dependency of a mechanical > > system on the current barometric pressure will be improved by an > > insistence on not having an origin somewhere near 1000 mbar? > Yes. How, precisely? Still, at least the variations in pressure (of order a few percent) will just be visible, if unclearly. Try plotting variations in the cosmic microwave background with an origin at 0K (which is unobtainable but never mind). Variations of one part in 100,000 are not going to show. And if you consider that by being so small, such variations are unimportant, please explain how anything would exist without them. -- Robin Stevens <rejs(a)cynic.org.uk> ---- http://www.cynic.org.uk/ ---- |