You misunderstand the statistics. A difference can be statistically significant with a trivial effect size, as here. p measures significance, r measures effect size. Here the numbers indicate that the difference is reliable (though p=.05 isn't significant where I come from especially in view of the huge sample size) but negligible.
He's not "Dr" Wakefield. He's just Andrew Wakefield, fraudster and disgrace to science. From the article: "On 24 May 2010, the GMC panel found Wakefield guilty of serious professional misconduct on four counts of dishonesty and 12 involving the abuse of developmentally challenged children, and ordered that he be struck off the medical register."
Yes they know about the hidden partition. What's worse is that they can't prove it's there, but you can't prove it's not there. They will rubber-hose you until you give up the secret partition. If you don't actually have a secret partition to give up, they'll rubber-hose you forever.
A lot of people load their JQuery libraries or whatnot from a CDN. In fact I think that's the preferred behaviour. There are multiple CDNs so the list is a bit longer and more annoying than you'd think.
If someone has an axe to grind, it's easy to make graphs lie. Lying with inferential statistics is a bit more involved, so that gets held off until third year. (No joke, I had a research methods assignment that required me to formulate conclusions then generate a dataset to support them.) But when you legitimately want to understand what's in data, or explain it honestly to someone else, the right visualisation beats statistics every time. Humans are pattern detectors, graphs contain patterns, statistics are just numbers. If you can't see a finding in the visualisation, and you have to rely on a statistical test to demonstrate it, there's a high risk that it's not really there.
This is oddly reassuring. I'm a reasonably smart guy, a strong statistician and programmer, and I cannot follow formulas. They're not intuitive and I have to painstakingly work through and translate them as if they were a foreign alphabet. I sometimes wonder whether that's just me, but it seems not. I note that someone in TFA's comments posted a mathematical explanation to demonstrate how much clearer it made things, and at the first Greek letter I skipped to the next comment. No doubt everyone in the 'general audience' would do just the same.
Do you have evidence to support any of your opinions? People have been looking at the relationship between applicant selection methods and job performance for as long as statistical tests have been around, and there are no compellingly good methods. People think it's easy until we go to the data and see how their preferred methods worked out objectively over a decent number of hires. Have you done this? How many people have you hired, how was their performance measured, how did they work out?
That would be the whole thrust of the article actually (although the stuff about Google's process being ineffective is all from 2009 which makes it quite old news): Google thought that gauntlet interviewing using methods quite similar to what you're advocating was giving them the world's best employees, because it felt like something that would be effective... but measurement showed that it wasn't.
It's been known to organisational psychologists for ages that the least-worst method is to observe candidates' performance in the actual job environment or a good simulation of it for a week or two. From reading the article and some of the things linked from it, it seems like IT hirers have begun to work this out independently.
Even for people with standard-issue XX or XY sex chromosomes, the journey from that to phenotypical gender is about a six-stage process. Most people arrive at one of two endpoints, but that still leaves another 62 or so different bit-patterns for phenotypical gender, and as the article suggests the low-order bits can be flipped after birth. A write-only boolean field doesn't really do the job.
No doubt they would all simply pair off and each invoice the other half a million dollars for an executive backrub, the fair market value of which is established by what the other guy charges?
So for those of us who currently don't get the joke, but are always willing to learn something, what's the punchline? Is it that one gun looks scarier but they are functionally the same?
I used to have a huge honkin rubber protective case. A few months ago I decided to live a little, and my iPhone was stripped of its rubber outfit and screen protector. It's glorious. Thin, clear display, stylish. No cracks or scratches on either surface yet. One drop onto anything but carpet and it'll be history, but honestly a year's use of a wonderful piece of design beats three years' use of some rubber lump too fat to even fit into a pocket.
In my experience TomTom vets its Map Share corrections by just not approving them. As one f'rinstance, it took 3 years for them to correct an illegal turn on two very busy roads in Sydney, despite me and presumably umpteen others reporting it every damn time. A new bridge near my work took over a year to feature on their maps. So I guess you should say that their vetted corrections are 100% accurate.
Why, I would be delighted to give you a complaint about Drobo. Here's one I prepared last year:
I just want to take this opportunity to warn people off buying a Drobo and/or DroboShare. For those who haven't heard of it, Drobo is an impressive-sounding box that you slide drives in and out of as you like, and it manages all the RAID and defragging and whatnot.
The acute problem is that without warning they can screw themselves up and refuse to mount. When this happens you are completely hosed because nothing except a Drobo can read the Drobo disk format. You will need to format and restore from backup. What's that, you say? You bought the Drobo because it seemed like the safest way to keep 8TB of storage going, given that that's really hard to back up? Tough tits. Yes, I know "RAID is not a backup" etc, and I did have actual backups from which I restored my data, but the difference between the Drobo marketing and the Drobo reality is staggering, especially given the Drobo cost.
The chronic problems? Drobo is slow, glacially slow. For some reason the makers claim that it can be used as "primary" or working storage, which is a laugh since the access times are significantly slower than a plain USB2 drive which is already uncomfortably slow for serious work. It will drive you nuts. The management interface will drive you nuts. The way Drobo lies to the OS about its capacity and available space, so you have to use the management interface to find out this basic information, will drive you nuts. And you will live in fear that unmounting the device by any means other than the management software will corrupt your data unrecoverably. Gods forbid you knock the USB cable out by accident.
DroboShare is the additional expensive component that turns Drobo from a USB drive into a NAS. It has its own problems. The least of which are that once you switch to DroboShare, neither Time Machine nor Lightroom will let you use the Drobo storage.
I thought Drobo was a great idea, it sounded awesome when described, I think CK even had good things to say about it, but I was wrong. It was a very expensive disappointment. This is my story.
tl;dr: Drobo and DroboShare suck, do not buy. Or if you must, buy my one.
All of http://www.devwebsphere.com/personal/drobo/ is a tragicomedy really, he bought one in 2009 and quite liked it for the first week, then spent 24 months battling with every problem I described and more, and he still hasn't given up! Me, I nearly stroked out when Disk Utility said "won't mount, can't fix, it's dead", and I never trusted it again after that.
A lot of the big boys have turned to something like Omniture (http://www.omniture.com/) when they reach this point. You can't afford not to know what your visitors are doing, and an in-house solution is pretty bleh compared to what Omniture does. One of my customers is using Omniture to track what 15M visitors a day are up to, and they know exactly where their traffic is coming from and going to, and how that turns into revenue.
The other advantage of not doing it yourself is that then your visitor analytics suite talks properly to your email campaign manager, on-the-fly website customiser, CRM suite, etc. Then you're easily doing things like installing TouchClarity to automatically present content that this visitor is most likely to be interested in and pay for, or telling your emailer "send a flight/accommodation package offer to everyone who added a flight to the Bahamas to their carts 3-5 days ago but didn't convert, and bring the campaign results back to Omniture". This stuff is amazingly profitable because it meets visitor needs.
FreeBSD hands off a virtual dsp to all programs that ask for/dev/dsp, and then mixes the result in kernel. No need for horrid esd and smacking up all your programs to use it, sound from multiple programs at once just works.
You should know (and you probably do and were just repeating a silly epigram) that the word history has a Latin origin and the fact that it begins with a three-letter sequence his is completely non-significant. You strike me as the kind of person who might also complain about manual or manicure because they begin with man.
Back on topic, I'm all for the use of she as a generic pronoun. I switch between them as often as possible without being confusing to the reader. I dropped the whole ``but he has become the de-facto generic pronoun'' line of argument when I realised that it has the same kind of effect as pirate being used as a term for copyright infringers.
Hey listen, buddy! Some of us have diagrams, charts and equations we'd like to display. At the moment I have to do that by having HTML with hundreds of little graphics inlined everywhere I use a math symbol or anything else that's not text. I would give anything to go to SVG. So would readers of papers on my website, both humans and spiders.
Why, certainly. One analyses emails entering a computer system, the other analyses a huge database of private information about human beings. Email doesn't have rights, humans do.
We sold an ML1210 to a business client who wanted fast, cheap and good. He got "cheap" and "good", but because most of his print jobs are single pages he sure didn't get "fast". The ML1210 and the 1250 we put in to replace it will both take over 40 seconds to print the first page if they have to come out of power-saving mode to do it. If they're still warm it's a far more pleasant 16 seconds.
Actually I'd be interested to hear from anyone who's overcome this, even if it was by disabling power-saving.
Slashdot Mirror
You misunderstand the statistics. A difference can be statistically significant with a trivial effect size, as here. p measures significance, r measures effect size. Here the numbers indicate that the difference is reliable (though p=.05 isn't significant where I come from especially in view of the huge sample size) but negligible.
Since you seem like that kind of guy I refer you to http://en.wikipedia.org/wiki/Effect_size#Pearson_r_.28correlation.29 so you can take Cohen's word for it instead of the author's and mine.
http://www.amazon.com/Mind-Games-Guide-Inner-Space/dp/0440556341/ ?
He's not "Dr" Wakefield. He's just Andrew Wakefield, fraudster and disgrace to science. From the article: "On 24 May 2010, the GMC panel found Wakefield guilty of serious professional misconduct on four counts of dishonesty and 12 involving the abuse of developmentally challenged children, and ordered that he be struck off the medical register."
Yes they know about the hidden partition. What's worse is that they can't prove it's there, but you can't prove it's not there. They will rubber-hose you until you give up the secret partition. If you don't actually have a secret partition to give up, they'll rubber-hose you forever.
Why are you lot the only people in the world entitled to privacy?
A lot of people load their JQuery libraries or whatnot from a CDN. In fact I think that's the preferred behaviour. There are multiple CDNs so the list is a bit longer and more annoying than you'd think.
Some links for background:
http://encosia.com/3-reasons-why-you-should-let-google-host-jquery-for-you/
http://royal.pingdom.com/2012/07/24/best-cdn-for-jquery-in-2012/
If someone has an axe to grind, it's easy to make graphs lie. Lying with inferential statistics is a bit more involved, so that gets held off until third year. (No joke, I had a research methods assignment that required me to formulate conclusions then generate a dataset to support them.) But when you legitimately want to understand what's in data, or explain it honestly to someone else, the right visualisation beats statistics every time. Humans are pattern detectors, graphs contain patterns, statistics are just numbers. If you can't see a finding in the visualisation, and you have to rely on a statistical test to demonstrate it, there's a high risk that it's not really there.
This is oddly reassuring. I'm a reasonably smart guy, a strong statistician and programmer, and I cannot follow formulas. They're not intuitive and I have to painstakingly work through and translate them as if they were a foreign alphabet. I sometimes wonder whether that's just me, but it seems not. I note that someone in TFA's comments posted a mathematical explanation to demonstrate how much clearer it made things, and at the first Greek letter I skipped to the next comment. No doubt everyone in the 'general audience' would do just the same.
Radiohead played in Osaka in 2008. I flew from Australia to hear them, so I know all about inconvenient concert venues...
Do you have evidence to support any of your opinions? People have been looking at the relationship between applicant selection methods and job performance for as long as statistical tests have been around, and there are no compellingly good methods. People think it's easy until we go to the data and see how their preferred methods worked out objectively over a decent number of hires. Have you done this? How many people have you hired, how was their performance measured, how did they work out?
That would be the whole thrust of the article actually (although the stuff about Google's process being ineffective is all from 2009 which makes it quite old news): Google thought that gauntlet interviewing using methods quite similar to what you're advocating was giving them the world's best employees, because it felt like something that would be effective... but measurement showed that it wasn't.
It's been known to organisational psychologists for ages that the least-worst method is to observe candidates' performance in the actual job environment or a good simulation of it for a week or two. From reading the article and some of the things linked from it, it seems like IT hirers have begun to work this out independently.
I like it when people who make claims provide citations so here's a good entry point to the research: http://mavweb.mnsu.edu/howard/Schmidt%20and%20Hunter%201998%20Validity%20and%20Utility%20Psychological%20Bulletin.pdf
An indicator for M/F isn't recording anything much about genetic sex. If that's what you're setting out to do you'll need a much bigger box: http://en.wikipedia.org/wiki/Sex_chromosome_disorders
Even for people with standard-issue XX or XY sex chromosomes, the journey from that to phenotypical gender is about a six-stage process. Most people arrive at one of two endpoints, but that still leaves another 62 or so different bit-patterns for phenotypical gender, and as the article suggests the low-order bits can be flipped after birth. A write-only boolean field doesn't really do the job.
No doubt they would all simply pair off and each invoice the other half a million dollars for an executive backrub, the fair market value of which is established by what the other guy charges?
Check out http://learn.adafruit.com/reading-a-analog-in-and-controlling-audio-volume-with-the-raspberry-pi/overview which I happened to be reading about 5 minutes ago. I'm not familiar with the One-Wire protocol but you can certainly bit-bang good ole analog sensors on the Pi. In Python, even.
So for those of us who currently don't get the joke, but are always willing to learn something, what's the punchline? Is it that one gun looks scarier but they are functionally the same?
I used to have a huge honkin rubber protective case. A few months ago I decided to live a little, and my iPhone was stripped of its rubber outfit and screen protector. It's glorious. Thin, clear display, stylish. No cracks or scratches on either surface yet. One drop onto anything but carpet and it'll be history, but honestly a year's use of a wonderful piece of design beats three years' use of some rubber lump too fat to even fit into a pocket.
In my experience TomTom vets its Map Share corrections by just not approving them. As one f'rinstance, it took 3 years for them to correct an illegal turn on two very busy roads in Sydney, despite me and presumably umpteen others reporting it every damn time. A new bridge near my work took over a year to feature on their maps. So I guess you should say that their vetted corrections are 100% accurate.
and then
HTH.
The other advantage of not doing it yourself is that then your visitor analytics suite talks properly to your email campaign manager, on-the-fly website customiser, CRM suite, etc. Then you're easily doing things like installing TouchClarity to automatically present content that this visitor is most likely to be interested in and pay for, or telling your emailer "send a flight/accommodation package offer to everyone who added a flight to the Bahamas to their carts 3-5 days ago but didn't convert, and bring the campaign results back to Omniture". This stuff is amazingly profitable because it meets visitor needs.
FreeBSD hands off a virtual dsp to all programs that ask for /dev/dsp, and then mixes the result in kernel. No need for horrid esd and smacking up all your programs to use it, sound from multiple programs at once just works.
And yet there they (or their members) are, running mp3.com, emusic.com, ...
Back on topic, I'm all for the use of she as a generic pronoun. I switch between them as often as possible without being confusing to the reader. I dropped the whole ``but he has become the de-facto generic pronoun'' line of argument when I realised that it has the same kind of effect as pirate being used as a term for copyright infringers.
Hey listen, buddy! Some of us have diagrams, charts and equations we'd like to display. At the moment I have to do that by having HTML with hundreds of little graphics inlined everywhere I use a math symbol or anything else that's not text. I would give anything to go to SVG. So would readers of papers on my website, both humans and spiders.
A pity about your Flash-only site. According to the poster below you have music available for download, which I quite wanted to listen to.
Why, certainly. One analyses emails entering a computer system, the other analyses a huge database of private information about human beings. Email doesn't have rights, humans do.
Actually I'd be interested to hear from anyone who's overcome this, even if it was by disabling power-saving.