[Nuclear1]

Why do you want the rich to pay more, simply because they can?  Isn't that a bit arbitrary?

Because the government gets more money that way, and it doesn't hurt them nearly as much as they like to whine about it.

Government getting more money is nothing but an excuse for it to spend more money, often on earmarks and pork for vote-buying.

1.  Raise taxes on rich
2.  Revive the Line Item veto
3.  ????
4.  Enjoy surpluses from now on

[SpardaSon21]

I'll accept a line-item veto only if it applies solely to federal appropriations.

[Kosh]

Living beyond our means is exactly what got us into trouble in the first place though.

Do I need to remind you that the number one cause of bankruptcy in the US before the financial crisis was medical costs? For some that's definately true, they have been living beyond their means, but for many it is just because of inflation.

[SpardaSon21]

My father is currently during bankruptcies to earn our family's daily bread, and a fair amount of clients come in with $500k in debt paying for their two houses, boat, and presumably credit cards, with all that debt obtained thanks to easy credit and low lending standards.  Even if there hadn't been a recession, sooner or later the piper would have had to been paid.

I'll need to ask him how many clients of his have had trouble paying medical bills, but chances are it is much lower than those living beyond their means.

[Turambar]

Oh snap, Sparda has anecdotes!  Whatever shall the people who disagree with his viewpoint do??!

[Kosh]

Cite a summary of the report

Post Date: February 3, 2005

Nearly half of all Americans who file for bankruptcy do so because of medical expenses, according to a new study released jointly by researchers at Harvard Law School and Harvard Medical School this week. The study, which is based on surveys of 1,771 individuals filing for bankruptcy, is the first of its kind to gather extensive information on the correlation between medical conditions and expenses and bankruptcy.

[Liberator]

The trouble with surveys is that the surveyor can get pretty much whatever results they want if they are so inclined.  Polls/Surveys/insert"objective" scientific measure of something here are pretty well meaningless these days.  Now I would accept an summary of bankruptcy filings for, say 5000, or so, people over a given period during the last 2 years as meaningful data.  Even that is really too small a sampling to make a sound argument from.

Mildly off-topic:
The solution to people filing bankruptcy for medical bills is not to have Big Papa Fed come in and pay for everything.  It's to fix an obviously broken system so that people can afford health care again without having to auction off their belongings and mortgage they're house twice.

Slightly more on-topic:
Bankruptcy should be the last chance shotgun blast for when the debt zombie just won't die any other way.  I suspect if you were to look a little more closely at those filings, most of the filing parties did so before they explored any other options they might have had.

[Kosh]

Bankruptcy should be the last chance shotgun blast for when the debt zombie just won't die any other way.  I suspect if you were to look a little more closely at those filings, most of the filing parties did so before they explored any other options they might have had.

Bankruptcy originally was that, but under the bankruptcy reform act of 2005 the debt collection will continue no matter what either until you die or somehow pay off all of it.

[General Battuta]

The trouble with surveys is that the surveyor can get pretty much whatever results they want if they are so inclined.  Polls/Surveys/insert"objective" scientific measure of something here are pretty well meaningless these days.  Now I would accept an summary of bankruptcy filings for, say 5000, or so, people over a given period during the last 2 years as meaningful data.  Even that is really too small a sampling to make a sound argument from.

Statistics FAIL.

1771 is a massive sample and completely representative. Bumping N up to 5000 would not get you any significant changes in terms of representativity.

In fact, 200 people would have probably been big enough to make this sample representative. 1771 is overkill.

This can be mathematically proven.

[Bobboau]

he's talking about bias in the sampling process, not sufficient size of a sample.

[General Battuta]

he's talking about bias in the sampling process, not sufficient size of a sample.

If he is, he's doing a poor job of it - he talked about increasing the sample size as a way to resolve his complaint.

[Liberator]

Sorry I'm not a statistics nazi?

Also, since I'm not edumacated, explain to me how 1771 is a sufficient sample size for a population topping 300 million most of which do not live in the likely sampling area?

[Ghostavo]

Sorry I'm not a statistics nazi?

Also, since I'm not edumacated, explain to me how 1771 is a sufficient sample size for a population topping 300 million most of which do not live in the likely sampling area?

1771 people is not a sample from a population of 300 million, since (or at least I hope) there aren't 300 million people declaring bankruptcy.

And according to this, a sample size of 1771 is enough to have about 2.3% as a margin of error.

[Mika]

Statistics FAIL.

1771 is a massive sample and completely representative. Bumping N up to 5000 would not get you any significant changes in terms of representativity.

In fact, 200 people would have probably been big enough to make this sample representative. 1771 is overkill.

This can be mathematically proven.

Care to explain this a little bit better? I deal with uniform Monte-Carlo statistics every day when I want to get a simulated image with any sort of accuracy, I need tens of thousands of samples per channel, as the statistical error behaves as 1/sqrt(N). However, I don't see much connection between Physical simulation and this.

1771 samples out of how many bankcrupties?

[Bobboau]

well without knowing the standard deviation it's a little hard to tell for sure I'd think

[Qent]

The trouble with surveys is that the surveyor can get pretty much whatever results they want if they are so inclined.  Polls/Surveys/insert"objective" scientific measure of something here are pretty well meaningless these days.  Now I would accept an summary of bankruptcy filings for, say 5000, or so, people over a given period during the last 2 years as meaningful data.  Even that is really too small a sampling to make a sound argument from.

Statistics FAIL.

1771 is a massive sample and completely representative. Bumping N up to 5000 would not get you any significant changes in terms of representativity.

In fact, 200 people would have probably been big enough to make this sample representative. 1771 is overkill.

This can be mathematically proven.

Well there's no need to be rude about it. I'm sure that most people in the world would have said the same, and a few months ago I might have too.

[iamzack]

I think GB could have been a little more rude, actually. Lib's response demonstrates that.

[Scotty]

Lib's response demonstrates that he thought GB was being rude.  No more than that.  Saying GB could have been more rude with his original post when looking at the response to that post as evidence strikes me as... unsound (that word doesn't really describe it, but meh).

1771 is a massive sample and completely representative. Bumping N up to 5000 would not get you any significant changes in terms of representativity.

In fact, 200 people would have probably been big enough to make this sample representative. 1771 is overkill.

I believe the question directed at this is "how can we know this isn't a biased sample."

[General Battuta]

Lib's response demonstrates that he thought GB was being rude.  No more than that.  Saying GB could have been more rude with his original post when looking at the response to that post as evidence strikes me as... unsound (that word doesn't really describe it, but meh).

1771 is a massive sample and completely representative. Bumping N up to 5000 would not get you any significant changes in terms of representativity.

In fact, 200 people would have probably been big enough to make this sample representative. 1771 is overkill.

I believe the question directed at this is "how can we know this isn't a biased sample."

That's a separate issue from the one Liberator brought up. You'll need to find the original paper and look at the sampling methodology. Liberator explicitly targeted sample size and made the mistake of citing an arbitrary value (5000), clearly not having done any math to arrive there.

Sorry I'm not a statistics nazi?

Also, since I'm not edumacated, explain to me how 1771 is a sufficient sample size for a population topping 300 million most of which do not live in the likely sampling area?

No need to Godwin yourself. You made an extraordinary claim about how survey measures were meaningless. You were wrong - or, at least, you cited terrible reasons for your argument. I'm here to correct you.

The population being sampled is not 300 million. The population in question is 'people going bankrupt'.

You only need to draw enough samples from the population to get a representative sample. Assuming your population sampling is truly random, you will reach a reasonable degree of representativeness very quickly, even if the population is enormous. For a given effect, we could confidently construct a sample representative of everyone on earth using a logistically feasible number of people.

It's much more important to have good sampling design (which would be something you could reasonably critique, though you'd have trouble doing it on every study done, which is a requirement for your argument) than a large sample, since the benefits of sample size shrink rapidly.

[General Battuta]

I just read the original paper.

The research design used a true random sampling of all bankruptcy filings that year (118,308 of them.) The only possible source of sampling error I can see in the design is non-response bias. However, the researchers compared a sample of non-respondents to respondents on a number of dimensions and found no significant differences. That bias is therefore controlled for.

Sampling error for a sample of this size out of a population of 118,308 is less than 1%.

Any questions?

So let me quote Liberator for a second...

Even that is really too small a sampling to make a sound argument from.

Could you please present to me the chi-squared test or forward stepwise logistic regression analysis that you used to reach this conclusion?

Oh, and just to further address your objections -

They used cohorts from multiple years.

The sample was truly random, so geographic bias is impossible.

They intentionally set their detection thresholds for a medical bankruptcy high, just to further lower the chance of a false positive.