http://news.ycombinator.com/item?id=23938

Few people are in a position to donate $20 M to charity. Fewer still who are willing to give up what they worked so hard to make. This puts sivers in the top (some small number) percent of human population in important objectively measurable criteria.

Personally, I enjoy learning more about people I admire, flaws and all.

Look at the figures. The performance of Slicehost follows a sawtooth like pattern. The quantity standard deviation is useful because it quantifies what to expect. Plus or minus one standard deviation means that ~ 2/3 of the time you will fall in that range.

If you think about the problem a little bit, you might be more worried about the standard deviation of the standard deviation. This, in fact, would be a useful quantity, but hard to measure.

EDIT below this line ------- Several comments below have commented that SD is somehow less useful if it's "large" (or large relative to the mean, or whatever). The reason people think large SDs are indicative of a poor experiment is that in school lab classes one calculates the SD and call it the "error".

The standard deviation is a measure of spread, if it's large then the spread is large. Knowing the spread has value. In this case, under the parent's experimental conditions EC2's performance is more constant than that of slicehost's.

A fair critique of the blog posting is that the error on the standard deviation may be large, depending on the experimental conditions. It is _not_ a fair critique to say that the SD is too high to make a prediction, you just have larger performance spread. Note that the performance spread described is not necessarily "error". The spread is inherit to either the server (as implied by the article) or the method (in which case it is an error).

You are 100% on point.

The parent's poster's point on conformity is well taken. Still, "independent" researchers are viewed with suspicion because research is hard. As psranga points out, to make a long jump a researcher has to stay on top of the incremental steps. Unfortunately, for 99.9% of people, to do this they must be fluent with the literature and much more importantly talk with other researchers. No one works in a vacuum (and don't bother bringing up counter examples like einstein or newton).

Conformity is a big problem. It might arise when too many PhDs are hustling for a small pie. I'm not sure.

Do you have access to a good electronics lab? A lab + H&H might be sufficient. I learned 80% of my electronics this way, but the last 20% was learned from old crusty EEs. If you don't have access to a lab, you can build your own (oscilloscope, function generator, power supply) for a few hundred dollars. I don't know where you can find old crusty EEs.

Bayesian Logical Analysis Physical Sciences by Gregory

Gregory's book explains a lot more of the engineering (autocorrelations, step size jumping, etc..). Even better, it discusses how to perform model selection using a clever annealing technique. Though model selection may not be of interest to you.

ps - MacKay's book is my nightly reading, so I'm not dissing MacKay :)

I've had a very nebulous idea that's been percolating in my mind: software designed for small machine shops to organize their work. It never occurred to me to organize hobby folk. awesome.

As the above triva factoid points out, the standard deviation is an important summary statistic. More interestingly by using mean, variance (or sd), skew, and kurtosis, you can describe almost any centrally concentrated distribution. Even distribution with heavy tails.

I think what the OP meant is that most 3+ sigma results are not truly 3+ sigma, because most distributions in this world are not gaussian, but instead have large wings. SD is most useful when you know what the underlying distribution is. Currently it's more in fashion to communicate spread using confidence intervals because they presume less about the underlying distribution.

http://stratocat.com.ar/bases/41e.htm

https://www.galileoscope.org/gs/

She'll be able to do a variety of experiments on it, and if it turns out she actually enjoys looking through it, you can then buy her something more pricey.

If you live in the bay area: http://techshop.ws/

If it's relatively simple, you can find a CAM house that will build the part for you, but then it becomes expensive. This place has an FDM house (wikipedia for details): http://www.emachineshop.com

You can find cheap drafters on craigslist who can make a solid model of your part that you would ship to the machine shop.

good luck

Comments URL: https://news.ycombinator.com/item?id=444017

Points: 17

# Comments: 4

What is the value of stock in a private company? Do you receive dividends on the stock? You can't sell the stock to other people? The only time it's of value is in a liquidity event. Suppose Fog Creek's not planning one in the future, what does this stock do?

I've been intending to play with CUDA compuation for a while, this seems like a great way to begin. It's especially nice that the documentation seems helpful. I'd be curious to hear peoples experiences with pycuda.

If we're going to rely on climate models to define policy, we should understand what they're good for. You say that 0.5 m v^2 is a good enough approximation until we say otherwise, but the same statement doesn't apply to climate models. The models are complicated, nonlinear, and the scientists running them have pressure to produce certain results.

You're wrong. The free market comes from the board of directors who approve such the "heads I win, tails I win" package. The board was not forced by any non-free-market laws. Thus, these compensation packages show the free market working.