I don't think it really works like this. In practice, most studies are totally uninteresting and their only purpose in life is either to not ever be read or to seed more uninteresting studies. Nothing of value is lost if they're wrong (and probably they often are). The studies that do matter are replicated because they're interesting enough that other people try to use them. So if a study discovers an interesting new effect or develops an interesting new tool then other (good) researchers jump on to the badwagon. If the original study was wrong then it'll be obvious pretty quickly. In addition, within every field it's an open secret which high profile papers are actually bollocks. It's usually obvious by just reading them. These papers are generally not contested by others for political reasons, but they are ignored because the field knows them to be crap.

In Europe, yes, this is a useful comparison. In the US the train system is slow and crap and has poor coverage.

Yes but these gripes by *contributors*. You also see users behaving in this way. Users haven't shared these frustrations.

Nasty learning curve? Really? I don't know JS, but I needed something to do simple dynamic DOM manipulations for an in-house web page. Couldn't do it with CSS. JQuery was easy to learn and it got the job done.

The problem is that these nerves don't grow back when they're cut. It doesn't matter how they're cut, they just don't grow back.

This has been done in the past with monkeys: http://en.wikipedia.org/wiki/H...

Same with Wakefield, who was planning to launch a diagnostic kit based on his MMR bullshit.

Why do you say retractions are secret? They very much are not: the journal publishes a letter of retraction from the authors. Hence RetractionWatch can, erm, watch for them. They may not be as highly publicized as they should be, but they aren't secret.

I agree in principle, but I think you're being a little over the top. Most contributors (rich and poor) to today's scientific and engineering challenges work in an institute that has access to the publications they need. For those who don't, they can access most articles by typing "[ARTICLE NAME] PDF" into Google. This works surprisingly often. If it's not available, just e-mail the author for a copy. Authors want their work read and don't give a shit about the pay wall. The paywall might be there, but it's not really stopping anyone from getting what they need.

In our field we call "bean plot" a violin plot. I agree it's better than a box plot, but it's basically just a histogram. Beanplot or boxplot, I think it helps to overlay the jittered raw data. Even a box plot is far better than a bar chart (which is distressingly common and little better than a table).

They both can be deceptive if you read them incorrectly or they're designed to deceive. However, it's harder to deceive someone with a graph than with the results of stats test. I really think graphs lie less easily than statistics.

I don't think you even need to be pushing people to do Bayesian stats. You just need to force them to graph their data properly. In *a lot* of biological and social science sub-fields it's standard practice to show your raw data only in the form of a table and the results of stats tests only in the form of a table. They aren't used to looking at graphs and raw data. You can hide a lot of terrible stuff that way, like weird outliers. Things would likely improve immediately in these fields if they banned tables and forced researchers to produce box plots (ideally with overlaid jittered raw data), histograms, overlaid 95% confidence intervals corresponding to their stats tests, etc, etc.

Having seen some of these people work, it's clear that many of them never make these plots in the first place. All they do is look at lists of numbers in summary tables. They have no clue in the first place what their data really look like, and know good knowledge of how to properly analyse data and make graphs. Before they even teach stats to undergrads they should be making them learn to plot data and read graphs. It's obvious most of them can't even do that.

On average, reviewers have the same skill set as authors who will get accepted (since that is the pool they are taken from). If authors are getting it wrong then so will reviewers.

I have worked on the retina, as it happens. :) What the retina sends the cortex is information about the relative intensity of red/green or yellow/blue in light reflecting from surfaces. This light is heavily influenced by the illuminating light source. So much so that it's possible for, say, an apparently green surface to be reflecting mostly red light. Yet you see it as green. Up until visual area V4, neurons are reporting that the surface is not green but red. In V4 we first see "colour constant" cells, the activity of which relates to perception. The dress illusion surely has its explanation here and not in the retina. The phenomenon is called "colour constancy" and requires neurons that have access to large regions of the image at the same time. This doesn't occur in the retina but only later in cortex.

It's not so cut and dried, actually. A lot of colour vision requires processing in the cortex so there isn't necessarily a clear cut relationship between the number of cone classes and an animal's colour acuity. A great example is the mantis shrimp which has a large number of different cone classes yet has crap colour vision. I don't know what bird colour acuity is or how it compares to our own, but don't assume it's necessarily better because they are tetrachromats. For instance, the wikipedia says that pigeons are pentachromats but they may not have access to the fifth channel. Many birds also have colour oil droplets in front of some photoreceptors in order to further tune their range. In effect, this may give them more than 4 cone classes.