Assorted links

Comments

Which other fields have the problem described in #4? Has anyone done similar studies?

Probably 95% of social science fields have this problem. ;)

It happens quite a bit in biology and medicine. Quite honestly, a cutoff of 0.05 is pretty lame; if an effect is real and the experiment isn't expensive to do, the observed pvalue diminishes rapidly with increased sample size. If the effect is not real, or is so weak it might as well not be, the pvalue does not diminish so quickly.

In cases were experiments are wildly expensive and the prior probabilities are not low, 0.05 is sometimes too strict. But most of the time science would be better served by a cutoff of 0.001. It's easy to implicitly test twenty hypotheses and passively neglect the failures, but much harder to implicitly test a thousand.

I bet a lot of false results would be uncovered by checking out the experiments where a more compelling p-value COULD be easy to get, with little additional effort, if the magnitude of the effect were as large as was being claimed. In these cases, publishing with P > 0.001 is inherently suspicious.

How would you analyze historical data where (truly natural) experiments cannot by definition be carried out?

Or you could just ditch the NHST framework all together...

http://library.mpib-berlin.mpg.de/ft/gg/GG_Null_2004.pdf

Masicampo and Lalande said their findings pointed to the need to educate researchers about the proper interpretation of null hypothesis significance testing

That's totally irrelevant. It isn't that researchers are stupid they are merely unethical . It's not a question of education really.

2: I thought they were called "harems"?

2. America seems to have great allies. They really share all the there values. Oh I forget it´s about oil.

I would love to go to a city with only women! rrrrr

#4 has already been nailed by xkcd:

http://xkcd.com/882/

As with most things.

PREDICTION : If Saudi Arabia creates "Women Only" cities then there will be cases within 6 months of Saudi males dressing up as women to get access.

6 is interesting. Time for the racialists to switch from grasping at the heritability of "g" to the heritability of some sort of "personality quotient" (GQ, maybe), as a means of essentializing and naturalizing inequality?

IQ is very predictive of outcomes. One narrow study of Swedish elites does not change that.

My understanding is that no one denies vital importance or heritability of personality traits. The only reason g/IQ features prominently in these silly debates is that it is much easier to measure (and that it does "explain" a significant fraction of what is observed).

So? I wasn't making a claim regarding whether or not high IQ was predictive of outcomes.

#4 There is a deficit for 0.1 > p > 0.05 which are roughly just as significant. Where are these interesting results we're not hearing about! Questionable practices lead to science covering up barely insignificant results!

It's probably a wash. You should ignore single experiments anyway. The interesting result would be clusters of papers for the same experiment that show a bias towards p = 0.05. The publishing threshold bias just represents a rule of thumb. Reach 0.05 and you're closing in on significance. Publish. Particle physics papers have a bias of results that just reach 5 sigma. That's because we think 5 sigma is the threshold for particle discovery. The current Higgs paper is 5 sigma. After awhile, the Higgs will be 6, 7, 8 sigma ... but no one will publish those papers except the review of particle physics.

I bet results of people's random statistical calculations on the internet have a bias towards p = 0.05, too.

If shooting someone is considered an extreme answer then, yes, at least one culture does seem to have a taste for extreme answers.

2: will women be freed from the Islamic dress code at least in these cities?

If not, Alistair above has nailed it;-)

Who is going to drive in this city?

The article's a little vague in its terminology, but I get the impression that Saudi Arabia is actually planning to build women-only industrial zones in existing cities, rather than entirely new women-only cities. I wonder about the transportation situation. Will there be dedicated women-only bus lines for these new areas?

Comments for this post are closed