Search This Blog

Friday, 20 September 2019

Review of BAYESIAN PROBABILITY FOR BABIES



                                                                               

List Price £7.99 from CHRIS FERRIE (University of Technology Sydney and father of four happy children)

Chris's Wikipedia page lists many further similarly pitched publications, Please remember that he is a quantum physicist.


Chris Ferrie very sensibly follows the example posthumously set by Rev, Thomas Bayes (1763) and his still alive sidekick Rev Richard Price by not including a statement of Bayes theorem in his seminal manuscript at all. Neither did Blaise Pascal or Pierre Fermat during their seventeenth century development of conditional probability, though they might have actually heard of it. As an albeit neurodiverse septuagenarian I find Ferrie's constructions to be almost as tedious to decipher as Bayes's. However, given all of our recent advances in Galtonian eugenics. I'm sure that the babies of 2019 will do much better.





ANNOTATED SUMMARY


On page 1 there is a red ball

On page 2 there are six balls of different colour (pieces of candy) on a cookie. Yum!

On page 3 there are also six balls of different colour on a cookie Some cookies have candy!

Image result for cookie with six candies

On page 4 there is a cookie with no candies  Some don't

On page 5 there is a bite-shaped portion of a cookie, with no candy

The same bite shaped portion appears on page 6, together with the remainder of the cookie. There are six pieces of candy on the larger portion. Did it (the bite-shaped portion) come from a candy cookie?

The picture on page 6 is reproduced on page 7. Either it came from a candy cookie---

---or it didn't .What are the chances? On page 8 the same bite shaped portion is compared with the remainder of the cookie, but now with no candies,

On page 9 there is a blank cookie with three symmetrically placed bite-marks

On page 10 it is stated that The probability of a no-candy bite, given a no-candy cookie is 1, and an attempt is made to illustrate this statement for babies by a conditional probability statement like

Pr (A l B)=1.

but where A and B are replaced by a picture of the bite shaped portion and a picture of the entire cookie. Confusing for some!

On page 11, the picture on page 9 wtth three bite shaped portions is reproduced, but now with six candies, Two of the bite shaped portions contain two candies, but the third one is blank.

On page 12 it is asserted that the probability of a no-candy bite, given a candy cookie, is 1/3. The 'justification' for this conclusion would appear to be completely arbitrary, and indeed totally incorrect if the baby had taken a random bite around the edge of the cookie illustrated on page 11. Undaunted, the author illustrates his fabrication by a conditional probability statement of the form

Pr(Al C)=1/3

where A and C are represented pictorially.

On page 13 it is stated that

Pr (Al B) > P (A lC)

since 1 is greater than 1/3

On page 14 there is again a picture of a candyless cookie, split into a bite shaped portion, and the remainder of the cookie. This is followed by the assertion that

So the no-candy bite probably came from a non-candy cookie!

Maybe babies would be less confused if 'probably' was replaced by 'more likely'.

On pages 15 and 16 there are pictures of ten delicious cookies, nine with six candies, and the other one blank.

But what if we knew there were 10 cookies. and all had candy but one?

The same ten cookies are reproduced on pages 17 and 18, but now all bitten

Take a bite of each, There are 4 no-candy bites. 3 bites are from candy.cookies . 1 bite is from a no-candy cookie

The same well-biiten cookies are reproduced on page 19.

1/3 of the candy cookies have a no-candy bite

On page 20 it is asserted that

P(Cl A)-=3/4

The probability of a candy cookie with a no-candy bite is 3/4

On page 21 there is a picture of the ten unbitten cookies

This is the prior distribution of cookies

On page 22 there is a picture of four bitten cookies

This is the posterior distribution of cookies

AND THAT'S THE ENTIRE CABOODLE





No comments:

Post a Comment