送交者: xj 于 2005-5-12, 18:21:33:

In response to MIRROR's pseudo-reasoning on coin tossing example a couple days.

The post is an expansion of my draft post intended on XYS, but decided to go to my blog after some revisions.

It's short, about 500 words.

所有跟贴:

- Confusing blog, but interesting follow-up discussion. -
**mangolasi**(1242 bytes)*2005-5-13, 16:26:06*- common people don't know bayesian, martingale or whatever -
**xj**(69 bytes)*2005-5-13, 17:37:54*

- common people don't know bayesian, martingale or whatever -
- "law of small numbers": if the probability is low, it won't happen -
**008**(313 bytes)*2005-5-12, 21:02:46*- the point is common people won't correct their own assumption -
**xj**(65 bytes)*2005-5-12, 22:07:50*

- the point is common people won't correct their own assumption -
- I think you should add a key point here. -
**Enlighten**(395 bytes)*2005-5-12, 19:40:01*- the difference on the absolute number of heads are sure larger -
**xj**(392 bytes)*2005-5-12, 20:52:00*- In gambling, the random walk has an absorbing barrier -
**wasguru**(42 bytes)*2005-5-12, 21:06:26*- I am not sure about mathematics behind that -
**xj**(417 bytes)*2005-5-12, 22:06:09*- absorbing barrier is when all your money is lost -
**Enlighten**(372 bytes)*2005-5-13, 10:18:08*

- absorbing barrier is when all your money is lost -

- I am not sure about mathematics behind that -

- In gambling, the random walk has an absorbing barrier -

- the difference on the absolute number of heads are sure larger -
- The flaw of binomial trial argument is not in the hidden sample size -
**wasguru**(399 bytes)*2005-5-12, 18:54:50*- yeah, viewing N as an sample size is a little bit over stretch -
**xj**(309 bytes)*2005-5-12, 19:02:27*- viewing a continuous coin toss as binomial loses a lot of information -
**wasguru**(160 bytes)*2005-5-12, 19:16:25*

- viewing a continuous coin toss as binomial loses a lot of information -

- yeah, viewing N as an sample size is a little bit over stretch -
- When the results are not robust the stats speaks. -
**cornbug**(0 bytes)*2005-5-12, 18:39:06*- that's your misconception. if results are not robust, stats won't help you -
**xj**(0 bytes)*2005-5-12, 18:51:40*- Not robust, either No difference or likely significant by stats. -
**cornbug**(0 bytes)*2005-5-12, 22:16:01*

- Not robust, either No difference or likely significant by stats. -

- that's your misconception. if results are not robust, stats won't help you -