Cap'n Refsmmat

Yeah, I was performing my duty and breaking SFN while setting things up. Should be sorted out now.

Oh, I just realized what's going on. Sorry about that; that screwed up all registrations over the past day or two as well. Short version: we moved our email over to a new server (which the forums will move to soon), and IPB was trying to send email through the old server. I had to prod it a bit to get it to send email through the right system.

I try to build an area surveillance system which can detect the introduction of radioactive sources, by using a mobile gamma spectrometer and a statistical model which maps the natural background radiation and detects sudden changes.

Looks like the scheduled task system (which automatically sends out these emails regularly) crashed a while ago. I just repaired it and it should catch up shortly.

I use this when I should be working: https://addons.mozilla.org/en-US/firefox/addon/leechblock/ There are similar extensions for Chrome. Block yourself from all the interesting websites and you might find that work is the most appealing thing to do. For the first few weeks I found myself visiting progressively more and more boring websites as I blocked myself from the good ones, until eventually it just wasn't worth the effort and I got work done instead.

The topic views counter is updated every 3 hours or so.

I'm not sure this is an effective way of correcting it.

Fair enough. I'd be interested to see a comparison between a professional methodological or statistical reviewer and the average harried and overwhelmed academic reviewer.

Replication studies are very difficult, since many papers do not provide details of materials used and methods descriptions are often incomplete. Journals usually won't publish replications either, since they're not novel. I don't think it's just intentional fraud or sloppy reviewers. This study of reviewing quality was amusing, for instance, since it tried to train reviewers to spot errors. Reviewers, knowing they were being tested, then looked for errors; they missed most of them. They weren't errors that were signs of deliberate fraud but just sloppy research. On the other hand, if you take papers already published by researchers from prestigious departments and relabel them as being from the "Tri-Valley Center for Human Potential," they will be rejected for "serious methodological flaws" by the same journals that published them.

I'm currently waiting on reviews on a paper I submitted, and I want to emphasize the "timeliness" point. I submitted back in April, replied to reviews in August, and I'm still waiting on a reply from the reviewers. It's embarrassingly slow. This is from an IEEE technical journal. PLOS ONE and a bunch of medical journals, on the other hand, say they'll take about a month or less. BMJ says the first decision will only take two or three weeks. Another interesting point is the ineffectiveness of peer review. The Cochrane Collaboration says There have been a number of amusing studies where manuscripts are submitted with deliberate errors, which the reviewers usually miss. But I've never seen a viable alternative presented.

That's possible. Or they were collateral damage when we deleted very old, unused accounts, and got mangled when converting forum software.

http://www.youtube.com/watch?v=aX6XMIldkRU

I'd like to see an example, such as Pythagoras's theorem proved via Bayes's theorem.

WYSIWYG editors are traditionally a major area where browsers cannot agree on standards of operation. It's very difficult to make them work. Invision, in their infinite wisdom, decided that WYSIWYG is the way forward and made their own shiny new editor, because BBCode is lame and mashing buttons that don't work is infinitely more satisfying.

In many scientific fields (e.g. psychology, biology, chemistry), there is no clear parametric model for the system in question, and no likelihood function to write down. Alternately, the likelihood may be incredibly complex and involve quantities which are practically impossible to measure. How does one perform Bayesian inference in these fields?

They're all inferior impostors, clearly.

It looks like for $30 I can get a Big Red Button, box, buzzer and wiring so anyone who presses the button will get a surprise. I may have to leave this on my desk in our shared office...

Of course you can keep track -- just read the Systematic Cookbook Review: http://ajcn.nutrition.org/content/97/1/127.abstract If you get the full paper, most ingredients had studies saying they caused cancer, along with competing studies saying they reduced cancer risks. Except for bacon, which uniformly caused cancer. I should point out that the article was For anyone interested, the article full text is available: http://www.nejm.org/doi/full/10.1056/NEJMoa1307352#t=article

Where do I get one?

That's probably when the system is saving database backups. We tried to pick the time with the fewest visitors, but a few people are bound to suffer through it.

I've never seen a DDoS on SFN, though I have seen a couple unintentional DoSes from poorly coded bots that hammered the server with multiple requests per second until I had it drop their packets. It's pretty easy to tell if there's a DDoS -- I have graphs of all traffic that goes across the network interface.

How does the server feel now?

I presume you're talking about Slowloris. We're much less vulnerable to it because we use lighttpd, which doesn't mind having loads of concurrent connections. As for the slowness, it may be the same problem as last time -- another website running on the server has a database table which has grown excessively large, and it tries querying it regularly, slowing the server to a crawl as it sorts through a couple million rows to find the right one. We'll try to fix this and hopefully implement a permanent solution.

Nitpicking: the NSA has also intentionally weakened encryption software, so this is not always a good option. More broadly: Privacy is not about having anything to hide. To quote myself: Privacy's point is not to hide things. It's a protection against government power. Giving the government the power to read your email, tap your phone, and record your porn usage isn't bad simply because it's embarrassing. After all, the data will likely only be seen by a computer. But it gives the government enormous power to make decisions about you -- decisions about whether you may take a commercial airline flight, get a security clearance, or even get a job -- without your knowledge or consent, and without you knowing how they make the decisions. If they reach the wrong conclusion or take data out of context, there is no recourse. Pervasive surveillance also means that the little crimes we commit every day -- pirating a movie, forgetting to report something on our taxes, smoking marijuana (and posting about it on reddit) -- are revealed to the government. They won't bother prosecuting most of them, but they can always choose to prosecute, and if a government official decides to abuse their powers they have a ready-made list of trumped-up reasons to arrest or harass you. In short, a lack of privacy gives the government the power to be even less transparent in its decision-making, and gives it yet more power over its citizens. It's not a question of discovering your fetishes or being embarrassed, and we shouldn't act as though only those with something to hide have reason to fear the government. Read this: https://chronicle.com/article/Why-Privacy-Matters-Even-if/127461/

“To consult the statistician after an experiment is finished is often merely to ask him to conduct a post mortem examination. He can perhaps say what the experiment died of.” -- R.A. Fisher