tl;dr Yes I would modify your approach. You are proposing a solution that grossly oversimplifies the problem by making a huge assumption that rarely holds in real life. It's not even wrong

That is a huge assumption that will not be true. Simply dividing a group randomly does not make the raw results coming out the other end meaningful. Do the 50 people in group 1 have the same starting weights as group 2? Same disposable incomes? Same amount of free time? Same stress level at home? Same family history? What if people move out of the area or otherwise lose contact? If you select groups of 50 randomly you will almost certainly have different distributions of underlying factors that could all plausibly have an impact on compliance with the regimen and effect of a well-followed treatment.

Maybe if you have a sufficient budget you can increase the sample size so that you are more confident that the two groups overlap. But are you sure? You don't even want to look at inputs and instead look only at end results so you will never be sure. And even if you could increase the sample size, with a sufficiently high multidimensional problem (which is generally the most important kind of problem) you can never truly ensure equality between groups.

So what would I do? First, I would do a better job splitting up the groups. No I won't explain how I would do it but you can find plenty of information on good experimental design elsewhere. Second, even with careful experimental design I probably wouldn't get perfect overlap, so I would build a model to test the effect of intent to treat on treatment rates and then treatment on the end variable, and do poststratification on the model versus the population to evaluate the intent to treat. The difference between what real researchers do and what you propose is like the difference between shooting a bullet and throwing it.

Yes I have a better alternative, and I'll do the best to help you out in the length of a comment. Everyone who runs social science experiments nowadays knows that there are problems with interventions; namely, some people who are assigned to the intervention group will not do it (e.g., proper diet and exercise) and some people who are assigned to the control group will actually do the intervention (e.g. they will eat healthy and exercise even if you don't tell them to). Modern statistics deals with these issues pretty well, and is able to give us metrics about how effective the actual treatment is, how well people are able to follow advice, and how good the advice actually is.

To get up to speed, you can read a good text such as this. If you already have a good stats background, you can go straight to chapter 9 and read from there.

Halfway through the dieting example it became clear that the author is completely unaware of multiple regression techniques, instrumental variables, or bayesian analysis, let alone experiment design. One would expect at least a cursory literature search (or even google) before writing so much about what is effectively a solved problem. He probably invents his own sort methodologies, revolutionizes page ranking algorithms, and rolls his own cryptographic hashes too.

Any time you talk about hitting the end of the universe you are going to have to deal with arrow of time issues, like mbone points out (e.g. you can't unmix the cream in your coffee).

But your point about predictive models still holds

Yeah but in one of the universes there was a mathematician named 6ul6r and so 6 *is* between 2 and 3

Something like this is probably the best idea. Or simply put up a big sign that says "Expected time to exit is [ ] hours". You can update that sign however frequently you want. People will make their own decisions about their own time management. If they don't want to wait in line for 5 hours, then they won't get in line when the sign says you will wait 5 hours.You won't need to coordinate with organizers or anything, just watch and update.

I know the quote is apocryphal, but the topic it was supposedly said about was a competing idea for a system of governance. About who gets power and who gets disenfranchised, very much like the question of Prop 8.

It is precisely in those important questions that opposing voices must be heard, and be given the opportunity to zealously defend their cause, as though by defense counsel for a heinous criminal. Even the Catholic Church has a Devil's advocate who argues against someone being given sainthood. Allowing people the freedom to show their support for bad things is good not just because it ensures our freedom should we ever hold a position that is widely viewed as immoral, but because how we treat those who disagree with us says a lot about what kind of people we are.

We need more of this in our public discourse.

I don't think they are saying that the typical user will experience a 5 minute longer delay; rather, the typical user will experience a shorter wait time. If the entire system is 40% more efficient, then even if you have to wait a little longer because the cab goes a little out of your way, that should be offset by a lower wait time. *Some* users will experience higher wait + travel times, but as they say here, no more than a few minutes.

The reason that they bring it up is because there are certain kinds of optimizations that will optimize the heck out of the average value, but they make the worst value an order of magnitude worse. For example, imagine a new proposal that would keep all the cabs in densely populated downtown and never go uptown whenever there was anybody waiting downtown. If there was a big enough ratio of people downtown versus uptown, then the average trip length would get better under this proposal, but the worst waiting time (for the uptowners) would be terrible.

Having said that, part of the reason why people take cabs is to get exactly where they are going, as fast as they can. If they wanted to get approximately where they wanted to go, with more people onboard, they would take a subway (or even the bus).

Gold is worth as much as it is because the qualities that make gold useful for so many products also make it useful as a material for currency. A material used for currency should have the following properties: it can be made into various forms useful for trading (such as coins or bars), it does not oxidize or otherwise decompose over long periods of time, it is rare enough but not impossible to find, it is relatively dense so you can efficiently measure by weight and not volume, and it would be nice if it even looked nice and shiny and pretty. You would be hard pressed to invent a physical material that satisfies the use cases of a currency as well as gold does. So any material that had the qualities that make gold useful in so many ways would probably get used as a currency first and foremost, driving the price up beyond its use in other applications.

You only lose (real) wealth via inflation if 1) you hold currency directly, or 2) you hold assets that are notionally defined and inflation is not sufficiently account for in the notional definitions.

So if you hold fiat currency directly in a non-interest bearing account, or in cash under a mattress, then you get hurt by inflation (1). Or if you hold bonds, or a savings account (2), whose interest rate is not large enough to compensate for inflation (particularly unexpected inflation), you will lose real wealth.

But note that these are not as big a problem as most people think. Generally, we do not hold our wealth in currency (1). Although it works acceptably as a store of wealth (and very well as a transaction medium), the opportunity cost is usually so high it is a poor asset over a long period of time, even if there were no inflation. For example, unlike dollars, gold has not lost money to inflation over a long period of time. However, if you held your wealth in gold, you would not have invested it in real estate, or the stock market, or a local franchise, etc. and so the real comparison that should be made is not gold versus dollars but rather gold versus any of the other things you could have invested in.

And (2) is much less of a problem because typically when we write notional contracts we account (as best we can) for inflation. So bonds during periods where inflation is expected to be high will have high nominal returns, and during periods of low inflation they will have low nominal returns.

Most people will not hold their wealth in a currency, such as dollars or bitcoin or gold, but rather, use those currencies to make transactions (ie. purchase goods and services, or temporarily as they sell one asset and buy another), and put their wealth into productive assets. The value of productive assets is not dependent on the value of the currency that they are denominated in, but rather, in the short run, the simple supply and demand for that asset, and in the long run, on the share of the total economy (across all currencies and denominations) that the asset is able to command.

If in 20 years, if some company such as Google controls 50% of the total productive capacity of the entire globe, then it doesn't matter if you sell 1/1,000,000,000,000 of the company in dollars, euros, yuan, or bitcoin, the relative value of those currencies will be balanced to reflect 50% * 1/1,000,000,000,000 of the world's GDP.

To kill hft, put a 0.001% tax on every transaction especially the cancelled ones. And hft will move to another jurisdiction over the next nano second.

FTFY. But seriously, the cancelled trade issue that you point out is in some sense the real problem. The way things are structured now, I think that it is difficult to say exactly how being faster to the trade is somehow unfair or morally objectionable. However, that is not what HFT does. Rather, it pings all over the market to figure out where the market is, and takes advantage of 1) their speed and 2) their ability to cancel a trade to send out massive numbers of orders that are never intended to be filled, scooping up free options.

A more effective approach is the one listed in gp post. Discrete time auctions would maintain almost all of the liquidity in the markets today, without the unfairness of cancelling orders left and right.

There are plenty of non-frivolous reasons why ATMs should be upgradeable. Banking is highly regulated, and if tomorrow the FDIC, the FRB, the OCC, or the CFPB made a rule about ATMs that could not be easily reconfigured for then an OS upgrade might be required to be in compliance. And it is unlikely that any sufficiently large organization has no security breaches on their internal network. A good defense in depth strategy would almost certainly devote some resources to making sure that ATMs are secure, to reduce the headline risk they pose if nothing else. And a part of that is ensuring that they are up to date.