The thing I liked is she talks about why bad science happens, poor studies maintain the social  order they are more likely to be picked up.

“Only“. That’s right. The study failed to find any association even of heavy or binge drinking with preterm birth, until they plucked out one particular group of women: those who drank heavily in the first trimester, then stopped.

These researchers had to work damn hard to pick out a group for which alcohol intake seemed to make any difference at all – and in that group, the confidence interval extended to 1.01. The researchers are going to have to work quite a bit harder to convince me that this association is sufficiently distinguishable from cherry-picking, a chance result of analysing and re-analysing the data in a desperate search for something “statistically significant” – especially given the complete lack of so much as a trend in the other groups. The authors claim a trend – but I can’t see it.

2.	(used with a plural verb) individual facts, statistics, or items of information: These data represent the results of our analyses. Data are entered by terminal for immediate processing by the computer.

3.	(used with a singular verb) a body of facts; information: Additional data is available from the president of the firm.

See this, this is not data, it isn’t summaries or statistics either but we will get to that later. Data would be publishing the responses received from Craig’s list, stripped of identifying markers of course, I mean we are decent people right? Right? Now, having done that you would have Data, on the responses of people who use Craig’s list to seek submissive women, which is not, technically the BDSM community. These are complex statistically ideas, I know but you know, maybe, just maybe guys on Craig’s list aren’t a random sample.

The author is reporting the results of a self selecting  survey with open ended responses, which are hard to statistically analogize, there are ways to approach it, you could for example tag (anal, force, impact play, etc…)  the entries and then look at the frequency of those tags, like most methods of open ended question statistics this introduces a mighty whack of experimenter bias, so it is particularly  important that your data is available for review so that your tags can be confirmed and challenged.

So if it isn’t data is it just mislabeled, is it summaries or statistics?

NO, either of those descriptions would have suggested more than the author collecting the worst that the online dating world has to offer as evidence of what? Perhaps alternative sexuality works well in the SCA? Back it up, or don’t call it data.

realistic title: Gene suspected to contribute to trans male transsexuality.

The facts of the case, 44% of ftM men people carried one variant of the gene vs 31% of cis women, this while interesting given the sample of 49 ftM men so we are looking at 6 more trans men than expected carrying the gene.

Leaving aside the small sample size the problem isn’t that the study was floored, it seems reasonable, it is that the reporting was poor, and that poor reporting of the study, even if we where looking at a sample of 1000′s of trans men we are talking about a one third increase in this gene among trans men, a interesting part of complex biology, but not enough to make “women feeel like men”.

I have talked about this this before but I think it is worth repeating. There is a very good interveiw with one of the reseachers who is quoted in the above article here.

A couple of things jump out at me, I once implemented a heart-attack and stroke risk calculator based on number crunching from the Framingham Heart Study, which is a huge study looking at the risk factors of heart disease and stroke, it didn’t ask for BMI, it made no different to life time risk, they looked at it and then threw it out when it was deemed no more effective than someone shoe size, here are the things which do make a difference.

I suspect people in general are eating better, which doesn’t mean less, have a look at those risk factors, blood cholesterol and blood sugar are on there, in the last 40 years there has been a big push to get people to stop eating animal based fats like lard and start eating vegetable based fats like olive oil. Less people smoke that 40 years ago, we have a whole lot of reasons to be happy about our situation, we are healthier that we were and we should be celebrating this.

Could we please start paying attention to how good the numbers are? please? Instead of blaming people without evidence because they you know, don’t look like models.

Writing in Psychological Science, a team led by Ian Spence of the University of Toronto describes a test performed on people’s ability to spot unusual objects that appear in their field of vision. Success at spatial tasks like this often differs between the sexes (men are better at remembering and locating general landmarks; women are better at remembering and locating food), so the researchers were not surprised to discover a discrepancy between the two. The test asked people to identify an “odd man out” object in a briefly displayed field of two dozen otherwise identical objects. Men had a 68% success rate. Women had a 55% success rate
Had they left it at that, Dr Spence and his colleagues might have concluded that they had uncovered yet another evolved difference between the sexes, come up with a “Just So” story to explain it in terms of division of labour on the African savannah, and moved on. However, they did not leave it at that. Instead, they asked some of their volunteers to spend ten hours playing an action-packed, shoot-’em-up video game, called “Medal of Honour: Pacific Assault”. As a control, other volunteers were asked to play a decidedly non-action-packed puzzle game, called “Ballance”, for a similar time. Both sets were then asked to do the odd-man-out test again.

Among the Ballancers, there was no change in the ability to pick out the unusual. Among those who had played “Medal of Honour”, both sexes improved their performances.

That is not surprising, given the different natures of the games. However, the improvement in the women was greater than the improvement in the men—so much so that there was no longer a significant difference between the two.
link

I’ve always been terrible at directions couldn’t make up or down off a map to save myself.

But the thing is now I can, I got myself a delivery job which meant I needed to know where I was going, so I do now know. It took me a week to get those skills, one week of driving around my local area on the clock.

How many times have you written something of as “thats just what (wo)men” do you think that would stand up to good experimental data?

I suppect that I will be reviewing a lot of papers from the journal of “Journal of Spurious Correlations” Along with the “journal of failing to correct for, income, access to health care…. etc”

Eve Bender

Conclusions about antidepressant efficacy based on a meta-analysis of FDA clinical trials raise serious questions.

In late February some headlines heralded bad news for those being treated for depression: “Prozac does not work,” declared one headline from the UK Guardian, “and nor do similar drugs….”

The news was in reaction to a metaanalysis of trials of selective serotonin reuptake inhibitors submitted to the U.S. Food and Drug Administration (FDA) from 1987 to 1999. The report appeared in the February 26 PLOS Medicine.

“These findings suggest that, compared with placebo, the new-generation antidepressants do not produce clinically significant improvements in depression in patients who initially have moderate or even very severe depression,” the authors wrote, “but show significant effects for only the most depressed patients.”

………
In their editorial, Turner and Rosenthal pointed out that the criterion set by NICE is “problematic, because it transforms effect size, a continuous measure, into a yes or no measure, thereby suggesting that drug efficacy is either totally present or absent, even when comparing values as close together as 0.51 and 0.49.”

They also noted that the NICE criterion is not a definitive measure, but a value that could be problematic as a litmus test for drug efficacy.

In the article, Kirsch and his colleagues noted that drug efficacy did not change as a function of initial depression severity, whereas placebo efficacy decreased as initial depression severity increased. “Efficacy reaches clinical significance only in trials involving the most extremely depressed patients,” the authors wrote, “due to a decrease in the response to placebo rather than to an increase in the response to medication.”

[link]
See that, blink and you would miss it, unless you dig deeper into the data you might never realise the the change in effectiveness is due to the placebo, not the drugs. This happens all the time, while working as a statistician I have come to the realisation that it is easier to mislead a reader than it is to give them a true picture of what the data says, I honestly struggle to not mislead those who I write for.
I wish the media was more questioning, if I had my way every journalism student would take courses in which they would have to make misleading findings given data in the hope that they would start asking questions of they way that studies where produced, did the researchers consider income, age, education level…?  is the effect due to outliers? did the researchers try a non parametric methods to confirm the results?

Humans are too willing to believe a man in a white coat, and we need to stop, it’s time to step out of the 80′s open our eyes, it’s time to be blinded no more