О корреляции.
Dec. 27th, 2006 05:11 amТут давеча опять всплыла тема, что мол корреляция означает причинность. Классическая шутка на эту тему (98% людей, умерших от рака, при жизни ели огурцы) всем хорошо известна. Мне тут пришла в голову ещё одна, для более узкого круга.
Вот графичек, на которым красным (левая ось) отображён тоннаж бомб сброшенных американскими тяжёлыми бомбардировщиками на Германию, а чёрным (правая ось) индекс германского военного производства:

Сост. по Richard G. Davis Carl A. Spaatz and the Air War in Europe, Washington D.C., 1993, pp.652,653
и Промышленность Германии в период войны 1939-1945 гг., М.:1956 сс.262,263
Видно, что чем больше бомбят, тем больше производят. Когда бомбардировки пошли на спад, индекс военного производства тоже упал. Я из любопытства посчитал коэффициент корреляции, получилось 87%; довольно неплохо. То бишь, чтобы немцы больше работали их нужно больше бомбить. Так выходит...
Вот графичек, на которым красным (левая ось) отображён тоннаж бомб сброшенных американскими тяжёлыми бомбардировщиками на Германию, а чёрным (правая ось) индекс германского военного производства:
Сост. по Richard G. Davis Carl A. Spaatz and the Air War in Europe, Washington D.C., 1993, pp.652,653
и Промышленность Германии в период войны 1939-1945 гг., М.:1956 сс.262,263
Видно, что чем больше бомбят, тем больше производят. Когда бомбардировки пошли на спад, индекс военного производства тоже упал. Я из любопытства посчитал коэффициент корреляции, получилось 87%; довольно неплохо. То бишь, чтобы немцы больше работали их нужно больше бомбить. Так выходит...
Интересно другое
Date: 2006-12-27 10:10 am (UTC)Re: Интересно другое
Date: 2006-12-27 10:51 am (UTC)Re: Интересно другое
Date: 2006-12-27 10:56 am (UTC)Re: Интересно другое
Date: 2006-12-27 02:03 pm (UTC)Re: Интересно другое
Date: 2006-12-27 03:10 pm (UTC)Re: Интересно другое
Date: 2006-12-27 03:42 pm (UTC)> Стратегические бомбардировки били по промышленному потенциалу, а не по конечной продукции.
Re: Интересно другое
Date: 2006-12-27 04:43 pm (UTC)Re: Интересно другое
Date: 2006-12-27 06:28 pm (UTC)Re: Интересно другое
Date: 2006-12-27 07:14 pm (UTC)"Могло бы и вовсе не вырасти. Воздействие порождает противодействие."
Этот тезис мог иметь смысл, если бы на графике был выведен показатель ввода в строй новых предприятий ВПК. Тогда вполне размумно было бы предположить "размен" - одни предпр. уничтожаются, вместо них открываются другие, в более укромных местах. Но на графике конечная продукция, а на убыль уже произведенной стратегическая авиация влияет слабо, если влияет вообще. Дело, конечно, и в противодействии тоже, но не Б-17 и ланкастеров :-)
Re: Интересно другое
Date: 2006-12-28 08:18 am (UTC)2) Этот тезис мог иметь смысл, если бы на графике был выведен показатель ввода в строй новых предприятий ВПК
Какая тут может быть связь ? Даже общих слов нет.
Re: Интересно другое
Date: 2006-12-28 08:42 am (UTC)Re: Интересно другое
Date: 2006-12-28 01:28 pm (UTC)Re: Интересно другое
Date: 2006-12-28 01:41 pm (UTC)Re: Интересно другое
Date: 2007-01-13 06:44 am (UTC)Это скан из книги "The Strategic Air War Against Germany. 1939-1945". Книга представляет из себя публикацию британского аналога американской серии "Strategic Bombing Reports".
Re: Интересно другое
Date: 2007-01-16 05:17 pm (UTC)Re: Интересно другое
Date: 2007-01-19 09:08 am (UTC)The British Bombing Survey Unit has carried out a more detailed enquiry into the effects which area attacks on German towns had on production in each defined group of industries in the Reich, in which is included Germany proper, Austria and the Sudetenland, between April, 1943 and June, 1944 inclusive. In this period a total of about 223,000 short tons of bombs were dropped in raids on some 50 German cities.
In carrying out the analysis use was made of all the material collected by the United States Strategic Bombing Survey, by the Operational Research Section of R.A.F. Bomber Command, and by the British Bombing Survey Unit itself. The material available for study comprised:
(a) studies of the effects of air attacks on some twenty industrial towns;
(b) completed questionnaires relating to labour, production as measured in sales figures, and air-raid damage, which were sent to every firm of more than ten employees in eighty-one towns that had been bombed;
(c) local central statistics from a number of towns giving labour figures, details about bomb damage, etc.;
(d) various sets of central Reich statistics on labour and production;
(e) a number of contemporary documents for which Speer and other prominent figures in the administration of German war industry were responsible.
Out of all this mass of data, statistics adequate for detailed analysis were obtained for fourteen towns that were practically free from air attack throughout the fifteen months covered by the investigation, and for twenty-one industrial cities that were heavily bombed.
The fourteen "control" cities are well distributed geographically in western and central Germany, but were mostly smaller industrial towns. The twenty-one bombed cities were a representative sample of all cities that were attacked. They accounted for 16 per cent. of all German industrial workers, and 36 per cent. of all workers in towns that were bombed, while between them they received 40 per cent. (86,200 short tons) of the total load of high explosive and incendiary bombs aimed at all towns during the fifteen months concerned. Of this tonnage of bombs, however, only 22 per cent. fell within the town areas – 6 per cent. within the central and the heavily built-over residential areas, and only 5 per cent. on the industrial and commercial areas (including railways and docks). An analysis of the accuracy and material effectiveness of these attacks showed that, period by period, they were on the average neither significantly better nor significantly worse than attacks on any other towns.
The first step in the calculation of the effects of area attacks on production in the Reich was to determine production losses in the sample of twenty-one bombed towns. Second, with these determined losses as a basis, the total decline in production in all bombed cities was calculated, allowance being made both for the different weights of attack which each town received, and for variations in the accuracy and material effectiveness of the attacks over the whole period. [1] In addition, the results of these analyses have been extrapolated to obtain an idea of the effects on production of the attacks which took place in the latter half of 1944 and in the first four months of 1945.
In order to determine the production loss in the sample of twenty-one bombed cities, it was first necessary to discover, by reference to the fourteen unbombed towns, what normal changes were taking place over the period in the monthly productivity per worker in each group of the official Reich classification of industries. For this purpose the productivity per worker (in terms of sales [2] in Reichmarks) was calculated month by month from the returns sent in by each group of industries in the fourteen towns taken together. The results represent the effect of general influences on German production which were independent of air attack during the fifteen months studied.
Re: Интересно другое
Date: 2007-01-19 09:12 am (UTC)With this information, it was possible to proceed to the next step of calculating the level of production that would have been reached in the twenty-one towns if they had not been bombed. First the numbers of workers in each group of industries were determined month by month in the twenty-one towns taken together. [3] An estimate of potential production, month by month, was then made by multiplying the number of workers in each group by the corresponding productivity per worker as determined for the fourteen unbombed towns. This estimate was then adjusted to allow for inherent differences in productivity between the two sets of towns. To do this, it was first necessary to calculate the ratio between the productivities of the two sets of towns during 1942 and the first quarter of 1943, when there was relatively little bombing at all. A further small adjustment was finally applied to allow for the estimated effects of the light attacks on the twenty-one bombed towns which took place during this control period.
Actual production for each industrial group in each town was calculated month by month by multiplying the recorded production per worker, as given in the sample returns, by the estimated number of workers in each group of industries in the town. Total actual production for each group of industries was then summed for the twenty-one towns.
The results of all these calculations showed that over the period studied total production was rising in both the attacked and the unattacked towns. On the other hand the figures indicate that total production was 13.7 per cent. less than what was potentially possible, taking the changes in the productivity per worker in the control towns as a standard.
During the period considered, the percentage of total output that was classified as war production was rising in all towns. The amount by which it would have risen .in the bombed towns, had they not been attacked, was estimated from the corresponding rise in the control towns, allowance being made for differences between the two sets of towns, during 1942 and the first quarter of 1943, in the proportion of armaments to total production in each group of industries. The level which the output of armaments would have reached in the last nine months of 1943 and the first half of 1944 in the twenty-one towns, if they had not been bombed, was then obtained by multiplying total potential production in each group of industries by the above percentages. The difference between these estimates and the actual production of war goods in each industrial group as given in the sample returns gave the loss in production of war goods.
The results showed that the rise in war production proper was slightly steeper in the towns that were bombed than in the controls. This means that losses in the attacked towns tended to be absorbed by sections of industry responsible for less essential production. The loss of war production was, in fact, only 3 per cent. of total potential production. On the other hand, loss of war production stated as a percentage of the twenty-one towns' potential war production was 6 per cent. Furthermore, it is apparent that the average monthly losses of war production proper were less in the first half of 1944 than in the second half of 1943, a fact which suggests that with increasing experience of air attack, the Germans became more skilled in diverting the effects of air attack on to the-civilian sector of industry. The evidence is perfectly clear, too, that over the period in question there were no "cumulative" effects of bombing on war industry. [4]
Re: Интересно другое
Date: 2007-01-19 09:14 am (UTC)What this analysis shows is, quite simply, that production in general, and war production in particular, was on the increase in the twenty-one towns from the middle of 1943 onwards, and that the bombing which they experienced depressed potential total production by little more than 10 per cent. and potential war production by about 7 per cent. With the help of all these figures estimates can be made of losses in total and in war production in all German towns that were bombed during the fifteen months studied. These estimates are based primarily upon the figures derived from the detailed study of the fourteen control and twenty-one attacked towns referred to above, and take into account the following variables:
(a) the weight of attack on all German towns in different periods;
(b) the "efficiency" of attack in different periods, as given by acres of city centre, residential area, and industrial area destroyed per ton of bombs carried. (The efficiency of attack for 1942 is taken as 100. Analysis showed that efficiency of attack was lowest in 1942, highest in the second half of 1943, and that it fell almost to the 1942 level in the first half of 1944);
(c) the percentage of each group of industries represented in the towns attacked;
(d) the vulnerability of each industry, as indicated by the detailed study of the twenty-one towns;
(e) the relative vulnerability of war industries and consumer goods industries in different periods; and
(f) the lag in production loss which follows an attack.
The results of the calculations are shown in Table 27. The figures give much the same answer as those derived from the detailed study of the twenty-one towns. Apart from an indication, certainly spurious, of significant losses in war production in iron and metal processing, output in other branches of war industry was clearly not significantly affected by the mounting toll of physical destruction in German cities during the last nine months of 1943 and the first half of 1944.
[1] In effect, the index used was acres of city centre, heavily built-over residential area, and industrial area destroyed per ton of bombs dropped. The average value for 1942 is taken as the standard (=100).
[2] Careful study showed that the advantages far outweighed the disadvantages of using sales figures rather than any other index of production in estimating the effects which area attacks had on industry. They represent the war goods and consumer goods actually made available to the nation. On the other hand they do not necessarily give a complete picture of production, since over the period studied some output went into the building of stocks held in the factories themselves or in Government depots. The actual increase in factory stocks was from 2,607 to 3,406 million Reichmarks. This increase provided an inventory cushion which at first helped to off-set the shocks which the German economy received after October, 1944.
[3] The extent to which each group of classified industries was represented in the conjoint lives of these cities varied considerably. The lowest figure was 2.1 per cent. for the ceramics industry, which in the whole Reich comprised 92,000 workers, while the highest was 42 per cent. for a branch of the iron and steel industry, which throughout the Reich employed 530,000.
[4] The calculations reveal relatively high losses in the iron and metal processing industries (as opposed to industries producing finished iron or steel products). From other evidence it is known that these indirect estimates are spurious.
Re: Интересно другое
Date: 2007-01-17 04:10 pm (UTC)