PMarione
Site Admin
Joined: 26 Mar 2007
Posts: 883
|
 Posted: Sat Apr 19, 2008 11:07 pm Post subject: Beware your stats |
 |
|
I somewhere explained how Tom Wareham completely botched the stats of his book The Star Captains by making a capital mistake: wrong sampling.
| Quote: | It would not be feasible to investigate in detail the career of every officer who held the rank of post captain at any point between the years 1793 and 1815, but a random-sampling method can be used to provide an outline of the officer corps.
In this study this has been achieved by taking all of those post captains whose surnames begin with the letters D, E and F, and who were technically available for service between 1793 and 1815: ie those reaching post rank before 1815 and who were still alive or had not reached flag rank before 1793. |
You can't make random-sampling using surnames beginning with a letter (for example it completely obscures officers whose name started with Mac or O' thus downsampling Irish or Scots officers).
The right way to do it would have to take every 10th or 20th officer from the complete list.
In predictive stats, the goal is to try to find conclusions from a sample that can be extrapolated to a population.
The choice of the sample is then very tricky and important because that sample is supposed to be an exact representation of the complete population otherwise any conclusion is without any value (see how often political polls are completely wrong).
This is the case with the sample of Tom Wareham which is not representative.
It's sad to see so much energy spent in a study which could have been very interesting if the sample had been correct.
Mrs Cavell's thesis (http://www.ageofnelson.org/Article05.html) has many interesting aspects but her hypothese that during the Long Wars the RN saw an important increase of officers of aristocratic origins is completely wrong.
It's true that the Admiralty made a real effort to gentrify the body of officers but without a real success.
According to her data, 85 honourables became Lts between 1781 and 1814. What she didn't calculate is that during that period, 7,173 people were promoted making the honourables 1.2% of the total.
According my own calculation (that I believe more accurate) for the period
120 promoted Lts were sons of peers
50 sons of "hon"
111 sons of baronets
----
281 officers of aristocratic origins (3.9%)
from that total we can substract
12 sons of peers
12 sons of "hon"
25 sons of baronets
---
49 whose father was an officer RN, a far more potent incentive
that leave 232 aristocrats (3.2%)
Difficult to conclude that the aristocratic element was prominent.
She also calculates (from Mackenzie's Trafalgar Roll) that 16 out of 587 young gentlemen were of aristocratic origins (2.7%) vs 6 out of 963 (0.6%) at the battle of Quiberon Bay in 1759.
Problem is that it's totally impossible to calculate the number of young gentlemen onboard a ship for the period: they could be on the muster books in any rating from AB to Yeoman of the Sheets even if many were rated as Captain's Servants in 1759 and Volunteers First Class in 1805.
Sampling problems again.
Her mistakes come probably from Michael Lewis.
In his Social History of the Navy, he made some such mistakes that can be taken in textbooks today.
In his social background chapter, he uses data taken from Marshall and O'Byrne, 2,354 from Marshall and 3,500 from O'Byrne.
From those he finds 1,800 people for whom he can find a background and here he makes his deadly mistake: he calculate all his stats from these 1,800.
12% aristocrats, 27.4% landed gentry, etc.
Problem is that his sample is wrong: in the 1,800 he found, he had 99.9% of the aristocrats (they are very easy to identify) and he had to make his calculation on the total of 5,874 giving 3.7% far less than 12%.
From my own calculation on 7621 officers, aristocrats count for 6% and landed gentry for 8%.
Another big mistake he has done is in the geographic origins of the officers.
| Quote: | | The Ireland of that day was very rich in lesser landed gentlemen little else: and their pedigrees were often longer than their purses. They formed a sort of closed aristocracy, the gulf between themselves and the true peasantry being quite unbridgeable. This probably helps to explain why Ireland's quota of officers was so much smaller than Scotland's. Whereas the 'lower deck' Scot always had a sporting chance of winning through to the Quarter-deck, his opposite number in Ireland was too illiterate, poor and down-trodden to have any chance at all. And much the same holds on the commercial level. There were many industrious merchant and business families on the rise in Scotland, but in Ireland many fewer. So most of Ireland's 176 officers were sons of 'landed' folk, with a number of professional people and a few 'commercial' sons from Dublin. Few if any were 'lower deck'. Yet let no one think that any disparagement of the Irish as seamen is intended. |
Pure racist non-sense.
If there were less Irish officers it's simply because the Test Act kept the catholics out of public office including army and navy commissions, and apparently there was a bunch of catholics in Ireland.
From my own calculations, England gave 23 officers/10,000 men of its population, Scotland 20, Ireland 19 and Wales 9.
The most represented proportionally were people from the Channel Islands (35/10,000) and Man (26).
In conclusion, if you do statistics there are 3 rules: check your sample, check your sample, and check your sample.
Another danger is to correlate stuff that have no correlation.
But that's another story.
@+P |
|
alexlitandem
Joined: 27 Mar 2007
Posts: 129
|
| Posted: Tue May 13, 2008 5:47 pm Post subject: |
|
|
Patrick,
Sample composition. Sample stratification. Representative of what universe / population. Robust. Reliable. Random. Stratified. Sampling intervals. Confidence Limits. Yada yada yada.
Etc.,
And, still, garbage such as that to which you refer is regularly published and - worse - thereafter cited / referenced as `statistics show'.
Ignorant fools? Retards? [ God, I feel so much better for getting that off my chest ] Or merely just self-serving liars ?
We need a Poll. |
|
