Statisitcs - common fallacy

Is Kahneman wrong?

Discuss the language of the universe.

Statisitcs - common fallacy

I'm reading Thinking, fast and slow by Daniel Kahneman and it's excellent, but on p 115 he says:
Our predilection for causal thinking exposes us to serious mistakes in evaluating the randomness of truly random events. For an example, take the sex of six babies born in sequence at a hospital. The sequence of boys and girls is obviously random; the events are independent of each other, the number of boys and girls who were born in the hospital in the last few hours has no effect whatsoever on the sex of the next baby. Now con­sider three possible sequences:

BBBGGG
GGGGGG
BGBBGB

Are the sequences equally likely? The intuitive answer- "of course not!" -is false. Because the events are independent and because the outcomes B and G are (approximately) equally likely, then any possible sequence of six births is as likely as any other. Even now that you know this conclusion is true, it remains counterintuitive, because only the third sequence appears random. As expected, BGBBGB is judged much more likely than the other two se­quences.

I think he is confusing two popular fallacies with a statistical fact.

Fallacy 1. After six girls, a boy is more likely (or indeed, after 666 girls). This is obviously false.
Fallacy 2. BBGBGG is more random/more likely than GGGBBB or GBGBGB. This is less obviously false.

Fact. The sexes in a run of six are most likely to be 3G3B (in any order), then 2G4B or 2B4G, then 1G5B or 1B5G, with six girls or six boys the least likely. (The first child must be one sex - with one in 2000 intersexed, but we'll ignore that - it's one chance in two the next child will be the same sex, one in four there will be three in a row, one in 8 four in a row, and so on.) So in any run of six, the chances of all one sex are one in 2^5 or 32, and the chance of that sex being girls is half again, so the chance of GGGGGG is one in 2^6 or 64. Expressed differently, the chance of G is one in 2, of GG one in 4 (with equal chances of GB BG or BB) GGG one in 8 (vs GGB, GBB, GBG, BBB, BBG, BGG and BGB) etc.

So has Kahneman overstated his case, or have I missed something?
Last edited by Shuggy on Jul 06, 2012 11:43 pm, edited 1 time in total.
The Wero Shop - products to make you laugh and think

Shuggy

Posts: 448

Print view this post

Re: Statisitcs - common fallacy

Daniel Kahneman is correct. Each of the three sequences is equally likely.

However, if you grouped all the possible sequences into kinds of sequence, then some kinds of sequence would be more likely than others, because different kinds of sequences will correspond to different numbers of individual sequences.

e.g. How likely is it that all the children have the same gender (all B or all G)? There are only two individual sequences for which this is the case: BBBBBB and GGGGGG. On the other other hand, there are many more individual sequences for which it is not the case, so we can say that "all the same gender" is unlikely - but it would still be wrong to say that some particular sequence of the "all the same gender" kind (e.g. GGGGGG) is less likely than some particular sequence of the "not all the same gender" kind - which is the fallacy which he is discussing.
If I ever start making posts like "On the banning and partial banning of words!" then I view my life as less than worthless and I hope that my friends here would have a collection to pay for ninjas to be sent to my home to kill me*. (*=humanely)

Paul Almond

Name: Paul Almond
Posts: 1541

Country: United Kingdom
Print view this post

Re: Statisitcs - common fallacy

Kahnerman is indeed correct, he just is a lousy teacher. The fact that he confused YOU, proves that. His chosen way to explain his point is much more convoluted than it needs to be, and therefore is more confusing.

A better way to describe the situation, I think, would be to point out that there is no statistical difference between looking at INDIVIDUAL BIRTHS, and looking at SETS OF SIX INDIVIDUAL BIRTHS.

igorfrankensteen

Name: michael e munson
Posts: 2114
Age: 64

Country: United States
Print view this post

Re: Statistics - common fallacy

Ah, I think I have it.
The chance of GGGGGG is one in 64.
The chance of BBBGGG is also one in 64.
The chance of BGBBGB is also one in 64.
There are 64 possible sequences, so "any possible sequence of six births is as likely as any other." as he says.

Since it is a book about how we think, I should look further into how I found that so hard. I seized on the rarity of GGGGGG, and clumped the mixed-sex birth orders (to my credit, only into different sex-ratio clumps, not one big clump), when the rarity of each of the birth-orders was what was to be contrasted with GGGGGGG, not the relative frequency of different ratios.

And thanks for the compliment, Igor.
The Wero Shop - products to make you laugh and think

Shuggy

Posts: 448

Print view this post

Re: Statisitcs - common fallacy

Every sequence is equally probable.
Warning: the content of the post above may content inaccuracies, nonsense or insults to human intelligence. Read at your own risk.

Posts: 1405
Age: 22

Country: Italy
Print view this post

Re: Statisitcs - common fallacy

Well other than the sex ratio not being exactly 50:50...
http://www.ncbi.nlm.nih.gov/pmc/article ... =pmcentrez
In America the battle is between common cents distorted by profits and common sense distorted by prophets.

byofrcs
RS Donator

Name: Lincoln Phipps
Posts: 7906
Age: 54

Country: Tax, sleep, identity ?
Print view this post

Re: Statisitcs - common fallacy

byofrcs wrote:Well other than the sex ratio not being exactly 50:50...
http://www.ncbi.nlm.nih.gov/pmc/article ... =pmcentrez

The example clearly assumes equal probabilities of B and G.

Moridin

Posts: 810

Print view this post

Re: Statisitcs - common fallacy

And if you don't assume equal probabilities, Fallacy 1 is even more of a fallacy...

Fallacy 1. After six girls, a boy is more likely (or indeed, after 666 girls). This is obviously false.

After 666 girls in a row, you should be concluding not that a boy is more likely, but a girl is overwhelmingly more likely.

virphen

Posts: 7288

Print view this post

Re: Statisitcs - common fallacy

virphen wrote:After 666 girls in a row, you should be concluding not that a boy is more likely, but a girl is overwhelmingly more likely.

No, because the assumption is that we are dealing with a fair coin. Or a fair uterus, as it were.

Moridin

Posts: 810

Print view this post

Re: Statisitcs - common fallacy

Moridin wrote:
virphen wrote:After 666 girls in a row, you should be concluding not that a boy is more likely, but a girl is overwhelmingly more likely.

No, because the assumption is that we are dealing with a fair coin. Or a fair uterus, as it were.

virphen

Posts: 7288

Print view this post

Re: Statisitcs - common fallacy

Moridin wrote: Or a fair uterus, as it were.
Don't blame the uterus. You would mean fair sperm if we were assuming fairness.
The Wero Shop - products to make you laugh and think

Shuggy

Posts: 448

Print view this post

Re: Statistics - common fallacy

Shuggy wrote:Ah, I think I have it.
The chance of GGGGGG is one in 64.
The chance of BBBGGG is also one in 64.
The chance of BGBBGB is also one in 64.
There are 64 possible sequences, so "any possible sequence of six births is as likely as any other." as he says.

I should have spotted this earlier, since I had already noticed the fallacy in another context: I have long scoffed at people who pay fortunes for personalised car licence plates, especially for "rare" ones like "MADONNA" and "JAG". ('You paid how much to remind people what brand of car they're screwed to?") All are equally rare - unique, in fact. That's the whole point of having licence plates.
The Wero Shop - products to make you laugh and think

Shuggy

Posts: 448

Print view this post

Re: Statisitcs - common fallacy

Shuggy wrote:
Moridin wrote: Or a fair uterus, as it were.
Don't blame the uterus. You would mean fair sperm if we were assuming fairness.

That's right - human sex is determined by the SRY gene found in the Y chromosome (thus from sperm). So we're default female unless this gene is expressed.
In America the battle is between common cents distorted by profits and common sense distorted by prophets.

byofrcs
RS Donator

Name: Lincoln Phipps
Posts: 7906
Age: 54

Country: Tax, sleep, identity ?
Print view this post

Re: Statisitcs - common fallacy

I think you understand what I meant.

Moridin

Posts: 810

Print view this post

Re: Statisitcs - common fallacy

byofrcs wrote:That's right - human sex is determined by the SRY gene found in the Y chromosome (thus from sperm). So we're default female unless this gene is expressed.

Actually, it is possible to have the SRY gene and still be phenotypically female. You cannot ignore the biological context.

Moridin

Posts: 810

Print view this post

Re: Statisitcs - common fallacy

Yes I didn't want to mention all the bizarre options. I read a lot about the SRY when arguing with Muslims as they believe that sex determination was decided 40 days after conception....e.g.

It took a while to explain that the sex was determined at conception but expression would be visible at the 40 days if you looked at abortions. They seem to think the genetic stuff happens later but in reality the sex of the foetus is already determined at conception.

Methodical Naturalism - 1
Muslim Apologetics - 0
In America the battle is between common cents distorted by profits and common sense distorted by prophets.

byofrcs
RS Donator

Name: Lincoln Phipps
Posts: 7906
Age: 54

Country: Tax, sleep, identity ?
Print view this post

Re: Statisitcs - common fallacy

The genotypic sex.

Moridin

Posts: 810

Print view this post

Re: Statisitcs - common fallacy

One possible way to define genotypic sex.
susu
susu.exp

Posts: 1690

Print view this post

Re: Statisitcs - common fallacy

byofrcs wrote:Well other than the sex ratio not being exactly 50:50...
http://www.ncbi.nlm.nih.gov/pmc/article ... =pmcentrez

Negligible
Warning: the content of the post above may content inaccuracies, nonsense or insults to human intelligence. Read at your own risk.

Posts: 1405
Age: 22

Country: Italy
Print view this post

Re: Statisitcs - common fallacy

susu.exp wrote:One possible way to define genotypic sex.

What I am saying is that SRY region determines genotypic sex, but from thereon, it is much more complex.

Moridin

Posts: 810

Print view this post

Next