Forgot your password?

Comment: Re:nice work (Score 5, Insightful) 464

Reminds me of a conversation I had with a student about a dozen years ago. GPS was all shiny and new in the civilian world, and he was an ex Army Ranger. I thought he'd be really gung-ho about GPS, but he said he preferred a paper map. When I asked him his reasons, he said "A GPS unit with a bullet hole through it is a door stop. A map with a bullet hole through it is still a map."

Ever since then I've operated in the belief that robust technologies trump cool technologies.

Comment: Re:R... (Score 1) 142

Python libraries are simultaneously advantageous and disadvantageous. Yes, they give a lot of leverage to solving a broad variety of problems, but last I checked many of them remained available only in Python 2. The success of the Python library ecosystem has actually interfered with the adoption of Python 3.

Comment: Re:Fully autonomous cars won't be ubiquitos (Score 1) 301

Pilots have a joke about the cockpit of the future. It will have an experienced pilot, manual override controls, and a German shepherd. The pilot is there to take the controls if necessary, and the dog is trained to bite the pilot if s/he reaches for the controls, thus ensuring that the pilot will only reach for the controls in a real emergency. In a pinch, you could replace the dog with high voltage running through the controls, eliminating the need for walkies.

Comment: Re:Why not car company? (Score 1) 301

I was answering to your comment, which seemed to imply that they are never point-to-point, when the system as a whole often are. And for the on-demand thing: If it leaves every 10 minutes or so, that's close enough.

I'm not trying to imply anything, I want to proclaim it boldly! Trains are not point-to-point. They don't go from your door to a destination, and don't generally provide direct routes. They don't depart at your convenience, and in the US, at least, they don't depart every ten minutes. In fact, there is virtually no reliable subway/tram/train service throughout the US except for a relatively small number of cities - more on why below.

Even where trains are available, it's not what I meant by point-to-point. When I lived in Boston, I took the subways/trams everywhere. However, the system is laid out as spokes, and to go to some destinations I'd have to ride all the way in to city center and then out another spoke. It made what would have been a 10 minute trip by autonomous vehicle into an hour to hour-and-a-half trip (depending on how the trains were running that day) with transfers. I chose to commute to a summer job by bicycle, 20 minutes each way, rather than lose 3-4 hours out of each day by having to ride the spoke system, transfer to buses, and then still hike a kilometer from the nearest bus stop to the job. However, biking wouldn't have been viable in the winter, nor would it be viable now that I'm older and missing the cartilage and half the ligaments from one knee. But part of my argument here is that the public transport solution wouldn't be viable for me now either - hiking a kilometer on ice and snow both ways daily would now be both painful and dangerous for me.

But shure, if you live out on the contryside, it gets harder, especially during the night. However it's funny that you mention airports, as they are often quite well served by public transport. I have myself several times taken the airport express train or bus to catch an early flight.

The fact that one end of the system is tethered to high-volume points of interest doesn't help people who live nowhere near any of the routes that go there.

The US is very large compared to Europe, and most folks here don't have access to transport systems comparable to those in Europe. Building the infrastructure isn't viable for most of the country because we have much lower population density. For better or worse, these are the reasons why the US has a more car-oriented society. I think autonomous vehicles can potentially have a huge impact here via pooled usage, reducing the need for individual ownership while still providing the point-to-point and on-demand benefits that are currently only available with privately owned cars.

Comment: Re:Why not car company? (Score 1) 301

As CrimsonAvenger points out, not everybody lives near a train depot.

You've also completely ignored the "on-demand" part of the statement. Trains don't depart on your schedule. A train won't pick you up at your front door at 3AM because you have a 6AM flight and the nearest airport is in a city 100km away. A subscription or rental service for autonomous cars should be able to handle that with no sweat.

Comment: Re:Why not car company? (Score 1) 301

So if you don't bring it in for service like ever and the brakes fail or you haven't patched the car's software in 10 years, is it still their fault? If you want Google to own that liability, you can also expect Google to set demands.

I have no problem with that. It's an autonomous vehicle, it can drive itself to the nearest link for software updates or over to the repair depot at night. If repairs are going to take more than my normal downtime or a nighttime emergency comes up, a taxi should be able to come and get me.

I actually anticipate a day when most people don't own their own cars, they subscribe to an on-demand point-to-point transport service instead.

Comment: Re:I'm assuming here... (Score 1) 769

by Frequency Domain (#46862477) Attached to: The Koch Brothers Attack On Solar Energy

Buying influence in politics is bad enough without people trying to make scientific issues political.

Scientific issues are political. Anything with the potential to impact the health, wealth, and well-being of humans, or to sort them into "winners" and "losers", is inherently political. Science can do all of those things.

Comment: Re:What other variable were examined? (Score 1) 668

by Frequency Domain (#45177271) Attached to: A Ray of Hope For Americans and Scientific Literacy?

The reason the story is interesting to non-statisticians is because anti-Tea Party stereotypes are proven wrong.

No, they're not. Bad analysis => cannot draw conclusions either way.

I have focused on Simpson's paradox in this thread because somebody else brought up controlling for education level, but it's not the only problem I noticed. I don't have any desire to go into a deep technical discussion of p-values and their interpretation, but I'll leave you with the thought that even with purely random data a proportion of them will be below your "critical threshold" alpha due to sheer chance - by definition alpha is the false positive rate for classifying effects as significant. If you try out a whole bunch of models at random, some of them will meet the alpha threshold even though they're not actually significant. The Yale professor strongly inferred that this was his methodology - he took a data set gathered for other purposes and tried things out until he got an interesting "significant" result. The fact that it's a "controversial" result is getting him lots of media attention. In the long run he may or may not turn out to be right, but this isn't good science.

Bottom line, since the analysis was done improperly (in several ways), you can't actually draw conclusions either way.

Comment: Re:What other variable were examined? (Score 1) 668

by Frequency Domain (#45176507) Attached to: A Ray of Hope For Americans and Scientific Literacy?

What does this have to do with whether Tea Party sympathizers understand science or not? What is the precise benefit of controlling for years of education? Do we care whether Tea Party people have more or less science knowledge than non-Tea Party people with the same number of years of formal education? Why?

Because it is possible to have both of these statements be true at the same time: "Tea Party people on average know more science than non-Tea Party People" and "At every education level the Tea Party people know less about science than the non-Tea Party people". The correct conclusion would then be that if you want to know how much somebody knows about science, look at their education but then adjust downwards if they are Tea Party people. Note that I'm not saying that that is the case. I'm saying rather that we can't tell whether the Tea Party identification has a positive or negative effect because the stupid social scientist did an improper statistical analysis. The more dominant education is in determining the outcome, the more important it is that you take it into account when considering the impact of other factors.

Do you want to condemn the Tea Party as being above average, but less above average than some other group of people? Who? And why?

I'm not condemning anybody except the Yale professor. I'm just a professional statistician who refuses to get suckered into drawing a conclusion one way or the other at this point based on shoddy analysis, and I'm trying to alert other /.'ers to the fact that the impact could actually still go either way if what you and I both agree is a major factor, education level, is properly accounted for.

Comment: Re:What other variable were examined? (Score 1) 668

by Frequency Domain (#45175337) Attached to: A Ray of Hope For Americans and Scientific Literacy?

Notwithstanding Simpson's Paradox, I'm still pretty sure science knowledge mostly comes from education.

Then you're not understanding Simpson's Paradox - this is exactly why you want to control for education.

Consider something floating on the water. Its movement will be a vector sum of wind effects and current effects. Suppose you happen to get a sample where the wind and the current are in opposite directions, and you try to estimate the effect of wind only, i.e., you leave current out of your model. At a minimum you will underestimate the impact of the wind, and if the current happened to be dominant in a large proportion of your sample you might even draw the false conclusion that free-floating objects move in the opposite direction from the wind! That's Simpson's Paradox - by omitting an important effect, you can actually end up drawing a conclusion that is the opposite of the truth.

"Be *excellent* to each other." -- Bill, or Ted, in Bill and Ted's Excellent Adventure