Talk:Scientific prediction

We need an article on scientific predictions. How accurate are they? How can laymen use them to make everyday plans? How well can policymakers rely on them?

What if people don't understand ranges and error bars, and fix on a mean or mode, as if it were guaranteed? For example, a highly publicized 1997 prediction of a 49 foot flood on the Red River of the North. How many people would think the flood would be exactly 49 feet (not 48 or 50) and fail to plan for a possible 57 foot flood? "But the scientists said 49 feet! How was I supposed to know it might be 5 or 10 feet more than that?" 

Before I start another article like "Factual disputes" which was immediately afd'd which discouraged others from collaborating, I'd like to see if anyone else in the world is interested in this topic. --Uncle Ed 14:53, 2 June 2006 (UTC)

At the forum of the Geophysical Institute, University of Alaska Fairbanks, Larry Gedney wrote:
 * The sciences are littered with failed predictions, yet are punctuated with equally many good ones that have pointed us toward a better understanding of the universe.


 * I suggest that if you want to write this, you stick well clear of global warming or climate change, which you simply can't write neutrally about. Stick to predictions of... bridge strength (or is that engineering?) or the t-meson mass, or some such William M. Connolley 15:47, 2 June 2006 (UTC)

Doc, please:
 * 1) I don't "want to write this" - I want Wikipedians to write it collaboratively. I don't think I can do it by myself.
 * 2) Please don't talk about *me* but about the article I'm proposing, or at least the topic. --Uncle Ed 16:16, 2 June 2006 (UTC)


 * I would be interested in collaborating on this topic, and I have several comments to start:
 * First, in the interest of consistency the article should only be written for the natural sciences (physics, chemistry, etc.) and biology, and all social sciences should be left out of it. The article would be massive otherwise and there is a common thread in the natural sciences and biology (e.g. the Scientific Method) that would be very unifying.  This is not however to be POV toward natural sciences but rather to allow for another group of contributors to make an equally qualified seperate article on the mechanics of predictions in social science fields (such as psychology, criminology, etc. - not, by the way, my area of expertise). Next, by its very title this article is going to be hopelessly broad because there are several branches of natural sciences, and there are several structures within each discipline to which the word "prediction" could apply; namely, these would be hypotheses, theses, theories, and laws, and the individual models within each.  Next, there would have to be sections on each of the following: what motivates predictions in science, how predictions are systematically quantified, how uncertainty in a prediction is systematically quantified, how the different types of uncertainties (systematic, etc.) are quantified and through which statistical formularies (i.e. frequency statistics or Bayesian statistics) are they handeled, and why.  The article should also devote an appropriate level of content to how predictions are ratified (i.e. why peer-reviewed scientific journal articles submitted by scientists are the most trusted, and why non-peer-reviewed opinions of non-scientists are in general the least trusted), and the types and locations of these resources for public access.  And finally, a good section on how the general public can use all of this information should be included.  These are just a few ideas of many that would make this article a quality, educational, NPOV contribution to Wikipedia. Astrobayes 17:15, 21 June 2006 (UTC)

"prediction" vs "forecast"
I am quite interested in collaborating as well. I am trying to understand better how we as scientists use and abuse the terms "prediction" and "forecast". There is an interesting book by Robert Goodell Brown, called Smoothing, Forecasting and Prediction of Discrete Time Series (2004, ISBN 0486495922). I haven't gotten a hold of it yet but it may provide some interesting insights. For example, in synoptic meteorology, a weather forecast differs from the preciction of a weather event. In volcanology, forecasting volcanic eruption based on monitoring data, differs from a prediction of possible future eruptions based on stochastic probability. In medicine, there are similar problems but the distinction is often even more fuzzy. All of these disciplines work with both approaches:
 * time series data to catch an oncoming event before it hits (volcanic eruption, hurricane, and a common cold) based on changing observational time series data and "epidemiological" knowledge of how such events typically develop;
 * probabilities of re-occurrence independent (e.g., "every August", "every 200 to 250 years", "everytime air masses invert, which is twice a month in summer", ...).

Any thoughts? --Carboxen 01:02, 23 July 2006 (UTC)


 * Indeed, within our peer group we scientists become comfortable with "prediction" and "forecast" as well as "model" as colloquial expressions for extra- and interpolation. Outside of the scientific community however, this is and can be the source of much confusion.  Also, there is a push within the natural sciences for the incorporation (revival?) of the old discipline of Bayesian statistics as an augmentation of the more common frequency statistics.  This article should accurately reflect these items, however a great deal of care should be taken if such a section is written by a non-scientist because colloquialisms within a peer group can often be seen by those outside of that group as intentionally misleading.  However, this is not the case with interpolation and extrapolation.  This section should be even handed and NPOV. Cheers, Astrobayes 19:31, 22 August 2006 (UTC)

statistics, etc.== ==

Tho unfortunately not planning to join in, i suggest that good luck. DGG 05:05, 12 October 2006 (UTC)
 * subsidiary topics be placed elsewhere.
 * explaining the connection between statistic prediction and true vs false is where non-scientists have the most difficulty (at least, I don't think I have every manged to teach it properly in a class=article or two)
 * perhaps instead of or in addition to true/false one can think of 2 or more alternative predictions, one & only one of which must be true --eg not whether global warming will occur but what the earth temperature will be.
 * what is needed most is non-controversal examples, both actual and thought-experiments.
 * and remember the logical problem of "other things being equal".
 * &, as a biologist, don't forget about the need to predict the past, i.e, to predict what future discoveries about the past will show, e.g. we shall find no dragon fossils as approximately equal to there have never been dragons.